UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Contextual and temporal variability in large-scale functional network interactions underlying attention Dixon, Matthew Luke 2017

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2017_september_dixon_matthew.pdf [ 5.6MB ]
Metadata
JSON: 24-1.0354463.json
JSON-LD: 24-1.0354463-ld.json
RDF/XML (Pretty): 24-1.0354463-rdf.xml
RDF/JSON: 24-1.0354463-rdf.json
Turtle: 24-1.0354463-turtle.txt
N-Triples: 24-1.0354463-rdf-ntriples.txt
Original Record: 24-1.0354463-source.json
Full Text
24-1.0354463-fulltext.txt
Citation
24-1.0354463.ris

Full Text

  CONTEXTUAL AND TEMPORAL VARIABILITY IN LARGE-SCALE FUNCTIONAL NETWORK INTERACTIONS UNDERLYING ATTENTION by Matthew Luke Dixon Hon B.Sc., The University of Toronto, 2006 M.A., The University of British Columbia, 2011 A DISSERTATION SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY in The Faculty of Graduate and Postdoctoral Studies (Psychology) THE UNIVERSITY OF BRITISH COLUMBIA (Vancouver) August 2017 © Matthew Luke Dixon, 2017       ii  ABSTRACT Attentional mechanisms filter and constrain the flow of information processing so that only the most relevant interoceptive and exteroceptive signals are highlighted for further processing. A variety of brain networks play a role in different facets of attention, including the default network, dorsal attention network, salience network, and frontoparietal control network. The present research used functional magnetic resonance imaging (fMRI) and graph theoretic analyses to examine the extent to which interactions between these large-scale brain networks vary across time and different contexts. We addressed the following questions: (i) is there a fundamental competition between networks involved in attending to perceptual versus conceptual information? (ii) is the frontoparietal control network―the key network implicated in the deliberate control of attention―a domain general system, or does it exhibit a finer level of organization related to perceptual versus conceptual attention? and (iii) how does network configuration vary during different modes of internally-directed attention characterized by conceptual elaboration versus interoceptive awareness? Our findings provide novel insights into these fundamental questions, and provide evidence that network organization dynamically changes across time and context. These findings caution against using resting state data to make general inferences about brain organization.    iii  LAY SUMMARY Attention is the process by which we focus on the most relevant information in any given moment. Prior research has identified a variety of brain networks play a role in different facets of attention. The present research used functional magnetic resonance imaging (fMRI) and graph theoretic analyses to examine the extent to which interactions between these brain networks change across time and different contexts. We addressed the following questions: (i) is there a fundamental competition between networks involved in attending to perceptual versus conceptual information? (ii) does the frontoparietal control network have a domain general organization, or is it composed of separate subsystems involved in regulating perceptual and conceptual information? (iii) how does brain network configuration change when individuals pay attention to their thoughts compared to their bodily sensations? Our findings provide novel insights into these fundamental questions, and reveal that brain network interactions changes across time and context.    iv  PREFACE I prepared the content of this dissertation. The introductory chapter includes content from a published review on different types of attention (Dixon et al., 2014, A framework for understanding the relationship between externally and internally directed cognition. Neuropsychologia, 62, 321-330), as well a published review on the frontoparietal control network  (Dixon et al., 2017, Hierarchical organization of frontoparietal control networks underlying goal-directed behavior. M. Watanabe (editor). The prefrontal cortex as an executive, emotional, and social brain. Springer. 133-148). Chapter 2  is based on a previously published paper on the default and dorsal attention networks (Dixon et al., 2017, Interactions between the default network and dorsal attention network vary across default subsystems, time, and cognitive states. Neuroimage. 147, 632–649). For these works, I was the first author and main contributor. Chapter 3 and 4 are based on new unpublished data that I collected, analyzed, and wrote up under the guidance of my advisor Kalina Christoff. The research conducted in this study was approved by the UBC clinical research ethics board (certificate number: H14-03078).   v  TABLE OF CONTENTS ABSTRACT ...................................................................................................................................  ii LAY SUMMARY .........................................................................................................................  iii PREFACE ...................................................................................................................................... iv TABLE OF CONTENTS ................................................................................................................ v LIST OF TABLES ......................................................................................................................... xi LIST OF FIGURES ...................................................................................................................... xii ACKNOWLEDGMENTS ........................................................................................................... xiv CHAPTER 1 - INTRODUCTION ............................................................................................... 1 Attention as a multifaceted process ........................................................................................ 2 Non-invasive brain network mapping in humans ................................................................. 5 Temporal and contextual variability in network organization ............................................ 5 Graph theory ............................................................................................................................ 5  Key networks involved in attention ........................................................................................ 6 The dorsal attention network ................................................................................................. 6 The default network ............................................................................................................... 8 The salience networks ........................................................................................................... 9 The frontoparietal control network ...................................................................................... 10 Overview of the current research questions ........................................................................ 12 The relationship between perceptual and conceptual attention networks ........................... 12 The organization of the frontoparietal control network ...................................................... 13 Network configurations during different forms of internal attention .................................. 15 CHAPTER 2 - EXPERIMENT 1: INTERACTIONS BETWEEN THE DEFAULT NETWORK AND DORSAL ATTENTION NETWORK VARY ACROSS DEFAULT SUBSYSTEMS, TIME, AND COGNITIVE STATE .............................................................. 17 vi  Introduction ............................................................................................................................ 17 Methods ................................................................................................................................... 21 Study 1 Effect size meta-analysis ........................................................................................ 21 Study 2 Participants ............................................................................................................. 22 Experimental conditions ...................................................................................................... 23 MRI data acquisition ........................................................................................................... 24 Preprocessing ....................................................................................................................... 24 ROI definition ...................................................................................................................... 26 Subsystem analysis .............................................................................................................. 26 Seed-based voxel analysis ................................................................................................... 26 Similarity analysis ............................................................................................................... 27 Machine learning classification analysis ............................................................................. 27 Logistic regression analysis ................................................................................................. 28 Dynamic FC analysis ........................................................................................................... 29 Temporal co-evolution of network interactions ............................................................. 30 Results ..................................................................................................................................... 31 Effect sizes of functional connectivity between the DN and DAN ..................................... 31 Patterns of functional connectivity between the DAN and each DN subsystem ................ 34 Stability of DN-DAN functional connectivity across cognitive states ................................ 36 Self-reported experience ...................................................................................................... 41 Stability of DN-DAN functional connectivity across time ................................................. 42 Temporal co-evolution of large-scale network interactions ................................................ 44 Discussion ................................................................................................................................ 48 Are the DN and DAN anticorrelated? ................................................................................. 48 vii  Variable interactions between the DAN and DN subsystems ............................................. 50 Contextual variability of DN-DAN interactions ................................................................. 52 Temporal co-evolution of large-scale network interactions ................................................ 54 Limitations ........................................................................................................................... 56 Conclusions ......................................................................................................................... 57 CHAPTER 3 - EXPERIMENT 2: FRACTIONATING THE FRONTOPARIETAL CONTROL NETWORK BASED ON INTER-NETWORK CONNECTIVITY .................. 59 Introduction ............................................................................................................................ 59 Methods ................................................................................................................................... 62 Participants .......................................................................................................................... 62 Experimental task conditions .............................................................................................. 62 MRI data acquisition ........................................................................................................... 63 Preprocessing ....................................................................................................................... 63 Definition of networks and nodes ........................................................................................ 65 Network visualization .......................................................................................................... 65 Hierarchical clustering analysis ........................................................................................... 65 SVM classification analysis ................................................................................................ 66 Comparing mean between-network FC ............................................................................... 66 Replication analyses ............................................................................................................ 67 Seed-based voxel analysis ................................................................................................... 67 Meta-analytic co-activation maps ........................................................................................ 68  Task-related flexibility. ....................................................................................................... 68 viii  Dynamic FC analysis. .......................................................................................................... 69 Results ..................................................................................................................................... 70 Evidence for distinct FPCN subsystems.............................................................................. 70 Differential coupling patterns with the DN and DAN ......................................................... 72 Replication and generalizability of differential coupling patterns ...................................... 75 Dynamic evolution of differential coupling patterns and network clustering ..................... 78 FPCN fractionation and task-related flexibility .................................................................. 80 Are the aFPCN and pFPCN subsystems of the same network or extensions of the DN and DAN? ................................................................................................................................... 81 Discussion ................................................................................................................................ 82 Functional organization of the FPCN .................................................................................. 83 Dynamic FC and clustering ................................................................................................. 86 Relation to other models of executive control and frontoparietal organization .................. 87 Limitations ........................................................................................................................... 89 Conclusions ......................................................................................................................... 90 CHAPTER 4 - EXPERIMENT 3: DISTINCT BRAIN NETWORK CONFIGURATIONS DURING EVALUATION-BASED AND ACCEPTANCE-BASED INTROSPECTION .... 91 Introduction ............................................................................................................................ 91 Methods ................................................................................................................................... 93 Participants .......................................................................................................................... 93 Task conditions .................................................................................................................... 94 Self-reports .......................................................................................................................... 96 ix  MRI data acquisition ........................................................................................................... 96 Preprocessing ....................................................................................................................... 96 Network nodes ..................................................................................................................... 97 Graph analysis: global properties ........................................................................................ 98 Community detection .......................................................................................................... 99 Rich club analysis .............................................................................................................. 101 Temporal variability of FC patterns .................................................................................. 102 Statistical testing ................................................................................................................ 102 Brain networks................................................................................................................... 103 Results ................................................................................................................................... 105 Subjective reports .............................................................................................................. 105 Global brain network properties ........................................................................................ 106 Within-network communication ........................................................................................ 107 Flexible reconfiguration of between-network interactions ................................................ 109 Rich club ............................................................................................................................ 112 Temporal variability of FC patterns .................................................................................. 114 High versus low acceptors ................................................................................................. 114 Discussion .............................................................................................................................. 116 Enhanced coordination between frontoparietal and interoceptive networks ..................... 116 Enhanced coordination between DN subsystems related to thought variability ............... 118 Enhanced rich club structure ............................................................................................. 119 High versus low acceptors ................................................................................................. 120 Limitations ......................................................................................................................... 121 Conclusions ....................................................................................................................... 122 x  CHAPTER 5 - SUMMARY AND CONCLUSIONS ............................................................. 123 Summary of main findings .................................................................................................. 123 Network organization is flexible ......................................................................................... 124 Attention relies on distributed network configuration ..................................................... 126 Conclusions ........................................................................................................................... 128 REFERENCES .......................................................................................................................... 129 APPENDICES ........................................................................................................................... 159 Appendix A: Supplementary material for experiment 1 .................................................. 159 Appendix B: Supplementary material for experiment 2 .................................................. 172     xi  LIST OF TABLES Table 1. Effect size of DN-DAN functional connectivity across studies ..................................... 32 Table 2. Self-reported experience ................................................................................................. 42 Table 3. Subjective reports about introspective experience........................................................ 106   xii  LIST OF FIGURES Figure 1.  Effect size of DN-DAN functional connectivity across 20 studies .............................. 32 Figure 2.  Anticorrelation as a function of DN subsystem ........................................................... 35 Figure 3. Comparison of within- and across-context similarity of DN-DAN connectivity .......... 37 Figure 4.  Accuracy of SVM classifier in distinguishing pairs of contexts .................................. 38 Figure 5. Whole-brain seed based analyses .................................................................................. 40 Figure 6. Temporal variability in DN-DAN interactions during rest ........................................... 43 Figure 7. Temporal co-evolution of network interactions ............................................................ 46 Figure 8.  Schematic illustration of temporal co-evolution of network interactions .................... 47 Figure 9. Network ROIs and topology .......................................................................................... 71 Figure 10. Visualization of the network topology ........................................................................ 73 Figure 11. Differential FPCN subsystem coupling patterns ......................................................... 75 Figure 12. Mean function connectivity between the FPCN subsystems and the DN and DAN ... 77 Figure 13.  Meta-analytic coactivation contrasts .......................................................................... 77 Figure 14. Dynamic network interactions and clustering during the Stroop task ......................... 79 Figure 15. Task-related flexibility ................................................................................................ 81 Figure 16. Mean between-network FC in each condition ............................................................. 82 Figure 17. Global network metrics computed across a range of correlation thresholds ............. 107 Figure 18. Nodes and network topology during rest................................................................... 108 Figure 19. Mean clustering for each of the 15 networks ............................................................ 109 Figure 20. Reconfiguration of between-network coupling patterns ........................................... 111 xiii  Figure 21. Rich club structure during evaluation and acceptance .............................................. 113 Figure 22. Network properties in high acceptors and low acceptors during the acceptance condition ..................................................................................................................................... 115      xiv  ACKNOWLEDGMENTS This journey was made possible through the support of many people that I am blessed to have in my life. I would not be where I am today without the love and support of my amazing family. I am very grateful to my advisor, Kalina Christoff, for facilitating my research life with her open-minded and relaxed nature, creativity, ability to see big picture ideas, and respect and trust she has always given me as a young researcher. I thank my dissertation committee members, Drs. Rebecca Todd, Janet Werker, and Todd Woodward for their time, and for being incredible people and researchers that inspire me. A special thanks to Dr. Todd Handy, who probably doesn't know the positive influence he has had on my research career. Thanks to Kieran Fox for many fruitful and enjoyable collaborations. A special thanks to Jess Andrews-Hanna and Nathan Spreng for critical guidance in network analyses. Thanks to Manesh Girn and Cameron Parro for graciously assisting my work in many ways. And of course much love to Eve De Rosa and Adam Anderson for being my first mentors. I had a wonderful experience collecting data at the UBC MRI Research Centre, so I am very grateful to Trudy Harris, Linda James, Alex Mazur, and Laura Barlow. Finally, this work would not have been possible without the financial support provided by a grant from the Natural Sciences and Engineering Research Council of Canada (NSERC), as well as financial support from Kalina Christoff.   1  CHAPTER 1 - INTRODUCTION In every moment, an enormous amount of information impinges upon the central nervous system―information that must be transformed and synthesized into a coherent representation of self, world, and the relationship between the two. This task is made possible by attentional mechanisms that filter and constrain the flow of information (Desimone & Duncan, 1995). Only those signals that are deemed most relevant are prioritized for further processing. Thus, attentional mechanisms highlight aspects of the inner and outer world, funneling select interoceptive and exteroceptive signals into conscious awareness, shaping decision making, reasoning, emotion, and memory (Chun & Turk-Browne, 2007; Duncan, Chylinski, Mitchell, & Bhandari, 2017; Hare, Malmaud, & Rangel, 2011; Krajbich, Armel, & Rangel, 2010; Pessoa, 2005).   Over the past two decades, neuroimaging and electrophysiological experiments have revealed a wealth of insights about the neural basis of attention (Buschman & Kastner, 2015; Corbetta & Shulman, 2002; Desimone & Duncan, 1995; Duncan, 2013; Miller & Buschman, 2013; Serences & Yantis, 2006). The majority of this work has focused on identifying the roles of individual brain regions. While this has yielded many important discoveries, a complete description has remained elusive given the complexity of attention as a cognitive construct, and the complexity of the neural infrastructure. The brain is a complex network of interconnected regions (Bassett & Bullmore, 2006; Bullmore & Sporns, 2009; Fornito, Zalesky, & Bullmore, 2016; Hagmann et al., 2008; Honey et al., 2009; Morecraft, Geula, & Mesulam, 1993; Petersen & Sporns, 2015; Sporns, Chialvo, Kaiser, & Hilgetag, 2004; van den Heuvel & Sporns, 2013) that operate like an orchestra, forming coalitions based on synchronized activity fluctuations that dynamically evolve across time (Allen et al., 2014; Betzel et al., 2016; Deco, Jirsa, & McIntosh, 2011; Hutchison, Womelsdorf, Gati, et al., 2013; Zalesky et al., 2014). These synchronized activity fluctuations allow information to propagate from sensory systems through the processing hierarchy, forming increasingly complex representations. It is only through this coordinated activity across the neural landscape that complex mental functions emerge and enable flexible and adaptive responses to different aspects of our environment. Thus, it is critical to explicate the large-scale brain network properties that support attention.  2   Based on task-based activation patterns and resting state functional connectivity analyses, a number of brain networks have been identified that are relevant to different aspects of attention, including the dorsal attention network (DAN), default network (DN), frontoparietal control network (FPCN), and salience network (SN). It has become increasingly clear that the interactions between these networks play an important role in attentional functions (Fox et al., 2005; Kelly et al., 2008; Kucyi et al., 2016; Sridharan, Levitin, & Menon, 2008). The majority of work to date has examined network organization during rest, with the assumption that the observed network features reflect the intrinsic functioning of the brain and are invariant to cognitive state (Fox & Raichle, 2007). Researchers have been eager to use this information as a biomarker to detect aberrations in clinical conditions (e.g., ADHD). However, this is problematic because the assumption of invariance has yet to be thoroughly tested. Thus, discerning the extent to which network properties exhibit contextual and temporal variability is an important foundational step in understanding the neurocognitive basis of attention.   The research presented herein uses graph theory to characterize the nature of brain network properties relevant to different aspects of attention. We address three major questions: (i) is there a fundamental competition between networks involved in attending to perceptual versus conceptual information? (ii) does the key network implicated in the deliberate control of attention―the frontoparietal control network―exhibit a domain general organization, or does it exhibit a finer-level of organization related to perceptual versus conceptual attention? and (iii) how does network configuration vary during different modes of internally-directed attention characterized by conceptual elaboration versus interoceptive awareness? The answers to these questions will be able to guide future research seeking to understand normative changes in the neurocognitive basis of attention across the lifespan, and pathological changes in clinical populations. Attention as a multifaceted process Attention is often likened to a spotlight that highlights a particular object or location for preferential processing. While this seems intuitive based on subjective experience, Desimone and Duncan's (1995) influential “biased competition” model instead suggests that attention is an emergent result rather than cause of preferential processing. The idea is that objects compete for access to processing resources―due to capacity limitations―and this competition is resolved by 3  a variety of neural mechanisms that exert a biasing influence on information processing, giving a competitive advantage to objects that are currently relevant. An object can be relevant due to: (i) physical stimulus features that potently activate the intrinsic properties of sensory systems (e.g., objects that are bright or moving) (Itti & Koch, 2001); (ii) deliberate control based on current goals and task demands (e.g., deliberately searching for one's car keys on a cluttered table) (Buschman & Miller, 2007; Corbetta & Shulman, 2002; Kastner et al., 1999; Moran & Desimone, 1985); (iii) long-term memory (novelty or familiarity can enhance the relevance of an object depending on the context) (Chun & Jiang, 1998; Chun & Turk-Browne, 2007; Summerfield et al., 2006); and (iv) emotional significance (biologically meaningful stimuli are preferentially processed) (Anderson, 2005; Anderson & Phelps, 2001; Todd, Cunningham, Anderson, & Thompson, 2012; Todd, Talmi, et al., 2012; Vuilleumier, 2005). These mechanisms combine to determine the overall relevance or salience of objects (Desimone & Duncan, 1995; Serences & Yantis, 2006). The most salient objects receive enhanced processing (i.e., stronger neural representation) while processing related to irrelevant stimuli is  suppressed, thus resolving the competition for attention (Desimone & Duncan, 1995; Kastner, De Weerd, Desimone, & Ungerleider, 1998; Moran & Desimone, 1985). While most research on attention has focused on how these mechanisms assign competitive weights to different visual objects, we can also conceptualize these weights as acting on a broader level to determine the relative salience of perceptual information versus conceptual or interoceptive information. Given that salience is determined by a suite of distinct attentional mechanisms, this naturally suggests that attention can only be fully understood at the neurobiological level within a network framework (Morecraft et al., 1993; Serences & Yantis, 2006).  Non-invasive brain network mapping in humans In recent years, there has been a surge of interest in understanding the brain in terms of coordinated large-scale functional networks (Fox & Raichle, 2007; Smith et al., 2013; van den Heuvel & Hulshoff Pol, 2010). Much of this work has emerged from functional connectivity (FC) analyses, which offer a powerful, non-invasive tool for delineating the functional architecture of the human brain. FC analyses typically involve computing the correlation between temporal fluctuations in blood oxygenation level-dependent (BOLD) signal across distributed brain regions (Biswal, Yetkin, Haughton, & Hyde, 1995; Fox et al., 2005; Greicius, 4  Krasnow, Reiss, & Menon, 2003; Power et al., 2011; Smith et al., 2009; Yeo et al., 2011). These analyses have repeatedly shown that a given brain region will exhibit correlated activation with a very specific collection of other regions, consistent with the idea that they work together as a functional system. A multitude of distinct functional networks have been identified, and associated with different cognitive functions (e.g., visual processing, language, somatomotor, attention) (Biswal et al., 1995; Buckner et al., 2009; Dosenbach et al., 2007; Fair et al., 2008; Fox et al., 2006; Greicius et al., 2003; Smith et al., 2009; Vincent et al., 2008; Vincent et al., 2007).   The validity of the networks identified by FC is supported by the fact that they bear a close resemblance to known anatomical connectivity patterns (Honey et al., 2009; Margulies et al., 2009; Raichle, 2009; van den Heuvel & Hulshoff Pol, 2010; Van Dijk et al., 2010). While there is not a 1-to1 mapping, anatomical pathways provide a strong constraint on the strength and spatial topography of functional connectivity (Honey et al., 2009). Additionally, FC analyses focus on BOLD signal oscillations in lower frequency ranges (0.009 –0.1 Hz), distinct from frequencies related to respiratory (0.1–0.5 Hz) and cardiovascular (0.6–1.2 Hz) signals (Cole, Smith, & Beckmann, 2010). Indeed, sophisticated analyses have revealed that the FC analyses are likely identifying network characteristics that are driven by hemodynamics related to true neural activity rather than non-neural sources (Cole, Smith, et al., 2010). Supporting this idea, resting state networks have been linked to power modulations in specific frequency bands in EEG and MEG data, providing evidence that they likely originate in the temporal fluctuations of neural activity (de Pasquale et al., 2010; Mantini et al., 2007).     The data used for FC analyses are often acquired during a “resting state” scan, during which participants lie in the MRI scanner with their eyes closed and simply allow their thoughts to naturally wander. The predominant use of resting state data is a function of its simplicity to acquire, applicability to a wide range of participant populations, and the assumption that it offers a window into intrinsic brain dynamics. Some have argued, however, that rest should be treated just as any other task-evoked mental state, and not reified as a baseline that offers a privileged view into brain function (Buckner, Krienen, & Yeo, 2013; Morcom & Fletcher, 2007). This perspective is predicated upon the fact that neural fluctuations during the resting state have been linked to a variety of mental processes, including memory retrieval, future planning, and self-5  referential thinking (Andrews-Hanna, Reidler, Huang, & Buckner, 2010; Doucet et al., 2012; Gorgolewski et al., 2014).  Temporal and contextual variability in network organization Counter to the idea that rest offers a window into the brain's intrinsic network organization, there is growing evidence that network properties dynamically change across time and context. For example, using a sliding window approach, numerous studies have demonstrated that specific FC patterns form and dissolve on the order of tens of seconds, and in some cases, reveal patterns that diverge considerably from static measures of network configuration (i.e., patterns related to summary measures of 5-10 minutes of data) (Allen et al., 2014; Hutchison, Womelsdorf, Gati, et al., 2013; Mitra, Snyder, Blazey, & Raichle, 2015; Shine, Bissett, et al., 2016; Zalesky et al., 2014). Moreover, FC patterns change in different task contexts (Buckner et al., 2013; Fornito, Harrison, Zalesky, & Simons, 2012; Krienen, Yeo, & Buckner, 2014; Spreng et al., 2010). In fact, FC patterns shift to such a degree that machine learning classifiers can predict the current cognitive state of an individual (i.e., which task they are currently performing) with remarkable accuracy (Gonzalez-Castillo et al., 2015; Milazzo et al., 2014; Shirer et al., 2012). Thus, it is critical to examine FC in a variety of contexts and as it evolves across time in order to have a complete understanding of how ongoing cognitive processing relates to the brain's functional architecture (Turk-Browne, 2013).  Graph theory Network properties can be ascertained in a variety of ways. Initial work focused on seed-based analyses in which the activation timecourse from a seed region of interest (ROI) is used as a predictor in linear regression analyses with the activation timecourse of every other voxel in the brain serving as outcome variables in separate univariate analyses (Biswal et al., 1995; Fox et al., 2006; Fox et al., 2005; Greicius et al., 2003). Positively correlated voxels ostensibly represent parts of the brain that belong to the same network as the seed region, although this assumption does not always hold (Wig, Schlaggar, & Petersen, 2011). Beyond this simple univariate approach, recent work has increasingly taken advantage of graph theory―a branch of mathematics that offers tools for quantifying the properties of any interconnected system, including social networks, transportation routes, the internet, and the brain (Bassett & Bullmore, 6  2006; Bassett et al., 2014; Fornito et al., 2016; Petersen & Sporns, 2015; Rubinov & Sporns, 2010). The system is modeled as graph G(N,E) with a set of nodes (N) and edges (E) representing the connection strength between nodes. Graphs can be binary (denoting the presence or absence of edges) or weighted (denoting the strength of edges, usually operationalized as correlation strength). Graph theoretical analyses then summarize the topological structure of the system along a number of dimensions.   For example, the capacity for local specialized processing can be assessed with the clustering coefficient, which computes the number of nodes around node i that are also interconnected, forming a dense circuit that ostensibly facilitates specialized processing (Onnela, Saramaki, Kertesz, & Kaski, 2005). The system can also be described in terms of its modular structure, by creating a partition that separates nodes into modules or networks via maximizing the strength of intra-modular connections and minimizing the strength of inter-modular connections (Blondel, Guillaume, Lambiotte, & Lefebvre, 2008; Newman, 2004). Based on the modular decomposition of the system, the strength of FC within and across networks can be quantified. Additionally, the roles of specific nodes can be delineated. For example, “hub” nodes can be identified based on the extent to which their connections are distributed across multiple networks (Guimera & Nunes Amaral, 2005). These measures have been successfully applied to brain networks, yielding a quantitative description of network organization (Bassett & Bullmore, 2006; Bassett et al., 2014; Fornito et al., 2016; Petersen & Sporns, 2015; Rubinov & Sporns, 2010).   Key brain networks involved in attention Below, I briefly describe several networks that have been implicated in attention-related processes, and are examined in the current research experiments. The dorsal attention network The dorsal attention network (DAN) is a large-scale network involved in perceptual attention. Experimental tasks that require participants to selectively attend and respond to the location or identity of visual or auditory stimuli invariably activate the DAN (Corbetta & Shulman, 2002). The DAN includes the frontal eye-fields (FEFs), anterior intraparietal sulcus extending into the superior parietal lobule (aIPS/SPL), middle temporal region (area MT), and ventral precentral 7  cortex (PrCv) (Corbetta & Shulman, 2002; Fox et al., 2006; Fox et al., 2005; Yeo et al., 2011). These regions exhibit correlated fluctuations in BOLD signal during task and rest states, suggesting that they operate together as a functional network (Cole, Bassett, et al., 2014; Krienen et al., 2014; Power et al., 2011; Yeo et al., 2011). Studies have often examined perceptual attention with cueing paradigms. In these tasks, a trial often begins with a symbolic cue that predicts the location (left or right side of space) of an impending target stimulus (Buschman & Kastner, 2015; Corbetta & Shulman, 2002; Hopfinger, Buonocore, & Mangun, 2000; Kastner et al., 1999; Miller & Buschman, 2013; Posner, 2012). When participants use such cues to proactively shift attention in space, there is often a sustained elevation in DAN activation (Bisley & Goldberg, 2003; Corbetta, Kincade, & Shulman, 2002; Corbetta & Shulman, 2002; Gottlieb, Kusunoki, & Goldberg, 1998; Hopfinger et al., 2000; Kastner et al., 1999).   Moreover, electrophysiological recordings in non-human primates have revealed that neural activity recorded from regions of the DAN exhibits a correspondence with the locus of attention (Bisley & Goldberg, 2003; Gottlieb et al., 1998; Moore & Armstrong, 2003). The receptive fields of neurons in these regions exhibit a precise spatial tuning, and only respond when behaviorally-relevant stimuli are present within a specific location of space (Gottlieb et al., 1998). Furthermore, this spatially-selective firing provides top-down input to perceptual processing occurring in extrastriate visual regions, thus enhancing the representation of behaviorally relevant stimuli at the attended location (Kastner et al., 1999; Moore & Armstrong, 2003). Further consistent with a role in processing external perceptual information, the DAN is involved in eye movements (Corbetta et al., 1998).   While this network was initially conceptualized as being specifically involved in top-down control (Corbetta & Shulman, 2002), abundant evidence has also revealed engagement of the DAN during bottom-up attention, when salient visual information (e.g., a suddenly appearing object) is presented (Bisley & Goldberg, 2003; Buschman & Miller, 2007; Gottlieb et al., 1998). It may thus play a general role in selective attention to perceptual information. That is, the DAN may support a perceptual saliency map―a sparse representation of the sensory world populated with only behaviorally-relevant stimuli (Ptak, 2012).   8  The default network The default network (DN) was initially recognized as a collection of regions that exhibit relative deactivations during task performance; that is, stronger activation during rest periods or interstimulus intervals than during trial periods that require attending and responding to perceptual stimuli (Buckner, Andrews-Hanna, & Schacter, 2008; Gusnard & Raichle, 2001; Mazoyer et al., 2001; Shulman et al., 1997). Key components of the DN include the medial prefrontal cortex (MPFC), posterior cingulate/retrosplenial cortex (PCC/RSC), posterior inferior parietal lobule (pIPL), lateral temporal cortex, and the medial temporal lobe (MTL) including the hippocampus and parahippocampal gyrus. These regions exhibit high metabolism during the resting state (Raichle et al., 2001) and correlated fluctuations in BOLD signal suggesting that they operate together as a functional network (Cole, Bassett, et al., 2014; Krienen et al., 2014; Power et al., 2011; Yeo et al., 2011) (Andrews-Hanna, 2012; Andrews-Hanna, Reidler, Sepulcre, et al., 2010; Fox et al., 2005; Greicius et al., 2003; Raichle et al., 2001).   Given that rest is associated with unconstrained mental activity, an early perspective labeled the DN a “task-negative” network and proposed that its functions were antithetical to goal-directed mental activity (Fox et al., 2005). However, recent findings have provided compelling evidence against this view, showing that the DN is frequently activated by experimental tasks that require goal-directed mental activity, so long as those tasks draw upon self-relevant information or associative knowledge (Andrews-Hanna, Smallwood, & Spreng, 2014; Spreng, 2012). For example, the DN is activated when individuals: (i) make judgments about their personality attributes (D'Argembeau et al., 2005; D'Argembeau et al., 2010; Mitchell, Macrae, & Banaji, 2006; Ochsner et al., 2005; Schmitz & Johnson, 2007); (ii) reflect on their own or others' emotions (Gusnard, Akbudak, Shulman, & Raichle, 2001; Lane, Fink, Chau, & Dolan, 1997; Lee & Siegle, 2012); (iii) recollect autobiographical memories (e.g., (Andrews-Hanna, Saxe, & Yarkoni, 2014; Spreng, Mar, & Kim, 2009; Svoboda, McKinnon, & Levine, 2006); (iv) imagine future events (e.g., personal goals such as becoming a doctor) (D'Argembeau et al., 2010; Schacter, Addis, & Buckner, 2007; Spreng et al., 2010); (v) infer and reason about the mental states of others (Amodio & Frith, 2006; Andrews-Hanna, Saxe, et al., 2014; Fletcher et al., 1995; Gallagher et al., 2000; Jack et al., 2012; Mar, 2011); and (vi) engage in periods of spontaneous thought (Christoff, Gordon, et al., 2009; Christoff et al., 2016; Ellamil et al., 2016; 9  Fox et al., 2015). The DN is therefore thought to be preferentially involved when attention is directed internally (Vanhaudenhuyse et al., 2011). However, given that attention can be directed externally during some of these functions (e.g., an external stimulus may trigger mentalizing or memory retrieval), it may be more accurate to characterize the role of the DN with respect to abstract conceptual/associative thought rather than internal attention per se. Indeed, DN recruitment can occur even during demanding externally focused cognitive task performance when a significant shift in cognitive context is required (Crittenden, Mitchell, & Duncan, 2015). Thus, whereas the DAN plays a role in attending to perceptual information, the DN plays a role in attending to abstract conceptual/associative information that goes beyond concrete stimulus characteristics. This accords with the fact that the DN is spatially and connectionally remote from primary sensorimotor cortices (Margulies et al., 2016).       Subsequent investigations have refined the concept of the DN and suggest that it is  composed of three distinct sub-systems based on task-activation and clustering patterns (Andrews-Hanna, Reidler, Sepulcre, et al., 2010; Andrews-Hanna, Saxe, et al., 2014). According to this view, the DN is composed of: (1) a core subsystem involved in linking affective preferences to autobiographical details and constructing a personal narrative and sense of meaning; (2) a medial temporal lobe (MTL) subsystem involved in retrieving learned associations that support scene construction and spontaneous thought; and (3) a dorsomedial prefrontal cortex (DMPFC) subsystem involved in social cognition, mentalizing, and conceptual knowledge. This recent conceptualization of the DN is important for considering how the DN interacts with other networks, which has yet to be thoroughly examined. The salience network When attention is directed internally, it can be directed not only to thoughts, but also to viscerosomatic sensations arising from the body. The capacity to be aware of bodily sensations is known as interoception and has been linked to a number of brain regions including the insular cortex (Craig, 2002; Critchley et al., 2004; Farb, Segal, & Anderson, 2012). Information about temperature, tissue damage, metabolic processes, and visceral sensations are conveyed to the insula through ascending pathways originating in the lamina I spinothalamic tract and vagus nerve and travelling through a number of brainstem and subcortical structures such as the periacqueductal gray (PAG), ventromedial thalamus, amygdala, and hypothalamus (Craig, 2002; 10  Critchley & Harrison, 2013; Damasio & Carvalho, 2013). In 2007, Seeley and colleagues introduced the “salience network” (SN), which they mapped out based on a seed region placed in the anterior insula. Correlated voxels were located in a number of the aforementioned brainstem and subcortical regions, as well as in the mid cingulate cortex (Seeley et al., 2007). More recently, Yeo and colleagues (2011) performed a whole-brain network parcellation and found a network composed of similar cortical regions with the addition of the anterior temporoparietal junction. Notably, this network split into two networks at a fine-grained level of analysis (i.e., their 17-network parcellation). We refer to these two networks as SN1 and SN2, which include the anterior insula and mid-insula, respectively.   Although Seeley and colleagues (2007) emphasized a broad role for this network in detecting salient internal or external events, this interpretation does not differentiate the role of the SN from the roles of the DAN and DN. Based on the noted viscerosomatic signals reaching this network, a parsimonious interpretation is that its central role pertains to interoception (Craig, 2002). Consistent with this idea, insula activation increases when participants attend to their heart beat and breathing (Critchley et al., 2004; Farb et al., 2012). Additionally, Farb and colleagues (2012) found a transition along the rostro-caudal gradient of the insula, with posterior and mid insular regions exhibiting stronger activation while attending to one's breath, and the anterior insula exhibiting stronger activation during a working memory task. This suggests that the anterior insula (and SN1 more broadly), may combine viscerosomatic signals with other contextual information in service of goal-directed task performance (Dixon, Fox, & Christoff, 2014a). To summarize, the SN1 and SN2 play roles in interoception, and may be recruited in response to a wide range of stimuli, because important events invariably elicit changes in body state that prepare the individual to take action.  The frontoparietal control network The ability to maintain attention during a lecture, or flexibly shift between writing a report and answering emails, or plan several steps ahead during a chess match all require cognitive control―the capacity to deliberately guide thought and behavior based on goals, especially in the presence of distraction or competing responses (Desimone & Duncan, 1995; Duncan, 2013; Gollwitzer, 1999; Miller & Cohen, 2001; Miyake et al., 2000; Posner & Dehaene, 1994; Posner & DiGirolamo, 1998; Stuss & Knight, 2002). Cognitive control has been studied with a variety 11  of tasks (e.g., Stroop, Flanker, task-switching) that require individuals to deliberately guide their actions based on a set of rules held in working memory. These tasks frequently engage a set of regions collectively referred to as the frontoparietal control network (FPCN) or Multiple Demand (MD) system (Cole, Repovs, & Anticevic, 2014; Cole, Reynolds, et al., 2013; Cole & Schneider, 2007; Crittenden, Mitchell, & Duncan, 2016; Dixon, Andrews-Hanna, et al., 2017; Dixon, Girn, & Christoff, 2017; Dosenbach et al., 2007; Duncan, 2010; Mitchell et al., 2016; Spreng et al., 2010; Vincent et al., 2008). Key regions of this network include the lateral prefrontal cortex (LPFC), anterior inferior parietal lobule (aIPL), intraparietal sulcus (IPS), posterior middle temporal gyrus (pMTG), and pre-supplementary motor area (pre-SMA). These regions exhibit correlated fluctuations in BOLD signal during rest and tasks (Cole & Schneider, 2007; Dosenbach et al., 2007; Power et al., 2011; Yeo et al., 2011).   The FPCN flexibly represents a variety of task-relevant information, and exerts a top-down influence on other regions, guiding activation in accordance with current task demands (Buschman & Miller, 2007; Crowe et al., 2013; Desimone & Duncan, 1995; Dixon, Fox, & Christoff, 2014b; Duncan, 2013; Egner & Hirsch, 2005; Miller & Cohen, 2001; Tomita et al., 1999). Neurons in these regions exhibit dynamic coding properties, signaling any currently relevant information (Duncan, 2010; Freedman & Assad, 2006; Stokes et al., 2013), including rules, objects, actions, and expected outcomes (Bunge et al., 2003; Cole, Bagic, Kass, & Schneider, 2010; De Baene, Kuhn, & Brass, 2012; Dixon & Christoff, 2012; Duncan, 2010; Jimura, Locke, & Braver, 2010; MacDonald, Cohen, Stenger, & Carter, 2000; Miller & Cohen, 2001; Parro, Dixon, & Christoff, 2017; Stokes et al., 2013; Woolgar, Hampshire, Thompson, & Duncan, 2011). Moreover, the FPCN supports proactive control (i.e., anticipatory maintenance of task information) prior to a required response (Jimura et al., 2010). Together, this work suggests a key role in the deliberate control of attention (Buschman & Miller, 2007; Duncan, 2013). Indeed, recent work has demonstrated that FPCN regions including the inferior frontal junction play a causal top-down role in modulating perceptual attention (Baldauf & Desimone, 2014; Barcelo, Suwazono, & Knight, 2000; Bichot, Heard, DeGennaro, & Desimone, 2015; Hampshire, Thompson, Duncan, & Owen, 2009). For example, LPFC damage is associated with reduced activity in the extrastriate cortex and impairments in detecting visual targets (Barcelo et al., 2000). Additionally, low frequency synchrony across frontoparietal regions reflects top-down attentional control (Buschman & Miller, 2007). Notably, the FPCN exhibits flexible connectivity 12  patterns (Cole, Reynolds, et al., 2013) and positively couples with the DN in some cases when attention is directed to internal thought (Christoff, 2012; Fornito et al., 2012; Spreng et al., 2010). For example, one study found that the FPCN exhibited increased functional coupling with the DAN during a Tower of London planning task that required attention to perceptual information, whereas the FPCN exhibited increased functional coupling with the DN during an autobiographical planning task that required attention to internal thought (Spreng et al., 2010). These findings suggest that the FPCN plays a broad role in the deliberate regulation of attention (Dixon et al., 2014b; Duncan, 2013). Overview of the current research questions Below, I describe three important topics related to attention and how examining contextual and temporal variability in brain network interactions can provide new insights into unresolved questions. The relationship between perceptual and conceptual attention networks  Individuals are rarely able to exert laser-like focus on a task without getting distracted by thoughts of a vacation, or where to eat for dinner. Behavioral research has shown that periods of mind wandering are associated with errors in externally directed task performance (Allen et al., 2013; Kam & Handy, 2014; Smallwood, McSpadden, & Schooler, 2008; Smallwood & Schooler, 2006). Conversely, when external task demands increase, individuals report fewer internal thoughts (Smallwood & Schooler, 2006). This has led to the idea that there may be a fundamental competition between attending to the external perceptual information and internal conceptual thought. Corresponding with these findings, studies have reported anticorrelated activation in the resting state between the DAN which contributes to perceptual/external attention, and the DN which contributes to conceptual thought/internal attention (Fox et al., 2005; Fransson, 2005). Thus, it would appear that competitive dynamics observed at the cognitive level are mirrored by competitive dynamics at the neural network level. This network antagonism is believed to support effective task performance (Kelly et al., 2008). While it has long been known that objects within a neuron's receptive field engage in mutually suppressive, competitive dynamics (Desimone & Duncan, 1995; Kastner et al., 1998; Moran & Desimone, 13  1985), these data suggest an additional type of competitive dynamics at a higher level, between perceptual and conceptual processing.    However, one limitation of the aforementioned findings is that they examined behavior and network dynamics in a limited number of contexts. Thus, anticorrelations may not be intrinsic, but rather, due to the context in which the networks have been previously examined. To directly test whether perception and conceptual processing compete, it is necessary to use task conditions in which both are relevant and equally likely to be prioritized by attentional mechanisms. In fact, we have argued that there is likely to be minimal competition between perceptual and conceptual processing in many naturalistic contexts (Dixon et al., 2014b). For example, during a conversation we seem to effortlessly synthesize information arising from the perceptual world (e.g., facial expression, body posture) with information arising from the internal conceptual world (e.g., inferences about the thoughts and feelings of our conversation partner). This suggests that perception and conceptual processing should be able to co-exist at the cognitive level. Whether this is also true at the level of brain dynamics can be tested by examining DN-DAN interactions across a number of contexts. This was the central goal of experiment 1. We also examined how DN-DAN interactions vary across time. Given that attentional shifts may occur on a relatively fast timescale, we hypothesized that the DN and DAN may alternate between periods of anticorrelation (reflecting a competitive relationship between perceptual and conceptual processing) and periods of positive correlation (reflecting a cooperative relationship that allows for integration between perceptual and conceptual processing). We addressed this possibility with a dynamic sliding window FC analysis.  The organization of the frontoparietal control network   Considerable progress has been made in understanding the role of the FPCN in the deliberate regulation of attention, including the way in which it flexibly couples with systems involved in perceptual and conceptual processing (Spreng et al., 2010). However, there is disagreement about its functional organization. Some work suggests that it operates as a unified domain general system (Cole, Repovs, et al., 2014; Duncan, 2010, 2013). Other work suggests that there may be a finer-level organization with anterior and posterior parts of the FPCN exhibiting relative preferences for more abstract and more concrete information, respectively (Badre & D'Esposito, 2009; Christoff & Gabrieli, 2000; Christoff, Keramatian, et al., 2009; Dixon et al., 14  2014a; Koechlin & Summerfield, 2007; O'Reilly, Herd, & Pauli, 2010). Elucidating the organization of the FPCN would provide insight into the way in which deliberate control is instantiated by the brain.  In experiment 2, we examined this issue with a hypothesis-driven network approach. Prior work has demonstrated that the FPCN is extensively interconnected with the DN and DAN (Spreng et al., 2013). Based on the divergent functions of the DN and DAN related to abstract conceptual thought and perceptual processing, respectively, we hypothesized that these distinct processing streams may be carried forward and reflected in the organization of the FPCN. Specifically, we examined whether the FPCN is topographically organized, with separate anterior and posterior zones connecting to the DN and DAN. If observed, this would potentially suggest that the FPCN is composed of separate subsystems for the deliberate regulation of perceptual versus conceptual processing. After finding evidence supporting this hypothesis, we then examined how this FPCN fractionation relates to the task-related flexibility (i.e., contextual variability) noted in prior work (Cole, Reynolds, et al., 2013; Fornito et al., 2012; Spreng et al., 2010). If the fractionation is not evident in some contexts, this would suggest that it emerges to support certain tasks, but is not a fundamental property of brain organization. Conversely, if the fractionation persists across multiple task contexts, this would support the idea that it reflects the fundamental organization of the FPCN. We also examined whether the FPCN fractionation could provide new insight into the dynamic evolution of network states.  Specifically, we looked at variability across time in the strength of clustering within the DN and DAN, which provides an index of specialized processing capacity, and may relate to fluctuations across time in the salience of conceptual and perceptual information. If the FPCN fractionation is relevant for understanding how the FPCN interacts with (and ostensibly regulates) other networks, then we should find a relationship between the anterior FPCN subsystem and temporal variability in DN clustering, and a relationship between the posterior FPCN subsystem and temporal variability in DAN clustering. In addressing these questions, our aim was to provide new insight into the network level architecture underlying deliberate influences on perceptual and conceptual attention.    15  Network configurations during different forms of internal attention  Humans often engaged in task-free introspective states during which attention is oriented internally. By directing attention to different internal signals, these introspective states can have markedly distinct subjective qualities. The clinical literature has followed contemplative traditions in making a distinction between evaluation-based introspection and acceptance-based introspection (Campbell-Sills, Barlow, Brown, & Hofmann, 2006; Hayes et al., 2006; Kabat-Zinn, 2009; Segal, Williams, & Teasdale, 2013). Evaluation-based introspection is the default tendency of most individuals; thoughts and emotions are evaluated and elaborated upon within the context of a personal narrative that connects past, present, and future, and provides a sense of meaning and coherence to life experience (Farb et al., 2007). In contrast, the aim of acceptance-based introspection is to adopt an open and unconstrained mindset characterized by a non-judgmental observation of thoughts and feelings (Kabat-Zinn, 2009). Because there is no attempt to feel or think in a particular way, this allows space for novel thoughts and feelings to continuously arise and dissolve, and for a deeper experiencing of concrete viscerosomatic sensations occurring in the present moment. Thus, a shift in salience occurs during this mode of introspection away from elaborative conceptual thought towards spontaneous thoughts and interoceptive signals. This shift in attentional focus is thought to enhance well-being by minimizing cognitive reactivity and enhancing awareness of the veridical experience of the body (Campbell-Sills et al., 2006; Hayes et al., 2006; Kabat-Zinn, 2009; Segal et al., 2013).  Although well characterized at the cognitive level, little is known about network configuration during these different introspective contexts. By taking advantage of subjective reports, our aim in experiment 3 was to delineate changes in network organization across evaluation-based and acceptance-based introspection. We hypothesized that acceptance would be associated with stronger coupling between the FPCN and salience network reflecting the greater weight given to interoceptive sensations. Additionally, we predicted that an open and unconstrained mode of thinking would be linked to enhanced communication between the medial temporal lobe and core subsystems of the DN (DNMTL and DNCore) based on a recent model suggesting that the DNMTL promotes thought variability (Christoff et al., 2016). Moreover, we expected to observe greater variability in DNMTL-DNCore coupling strength across time, corresponding to the greater variability in thought content that presumably occurs during 16  acceptance when attention is not fixated on a particular thought stream. We also examined interactions between the DNCore and salience network to delineate the relationship between evaluative thought and interoceptive processing. During evaluation, the DNCore and the salience network may be positively correlated if the evaluative self-referential processing supported by the DNCore is grounded in the ongoing viscerosomatic signals registered by the salience network. Alternatively, a negative correlation between these networks would imply a dissociation between evaluative thought and bodily sensations. Finally, we examined global network properties to delineate the full extent of network reconfiguration during acceptance versus evaluation. Specifically, we used measures of clustering, global efficiency, degree, and rich club structure to examine network modularity and capacity for parallel information transfer and integration. The answer to these questions will provide valuable information about the composition of different introspective states. Further down the road this could potentially be used as a neurocognitive index to identify adaptive versus maladaptive introspective tendencies.    17  CHAPTER 2 - EXPERIMENT 1: INTERACTIONS BETWEEN THE DEFAULT NETWORK AND DORSAL ATTENTION NETWORK VARY ACROSS DEFAULT SUBSYSTEMS, TIME, AND COGNITIVE STATES  Introduction One of the most influential findings to emerge from network neuroscience is the demonstration of anticorrelated networks, ostensibly reflecting competing functions (Fox et al., 2005; see also Fransson, 2005; Golland, Golland, Bentin, & Malach, 2008). The default network (DN) is involved in a variety of internally-directed conceptual thought processes, including self-reflection, autobiographical memory, future event simulation, and spontaneous cognition (Andrews-Hanna, Smallwood, et al., 2014; Buckner et al., 2008; Christoff et al., 2016; Ellamil et al., 2016; Fox et al., 2015; Fox, Andrews-Hanna, & Christoff, 2016; Raichle, 2015; Raichle et al., 2001) and exhibits decreased activation during many cognitive tasks that demand external perceptual attention (Greicius et al., 2003; Gusnard & Raichle, 2001; Shulman et al., 1997). In contrast a collection of regions, known initially as the “task-positive” network, demonstrate activity increases during cognitive tasks that require externally focused visuospatial perceptual attention (Cole & Schneider, 2007; Corbetta & Shulman, 2002; Dosenbach et al., 2006; Duncan, 2010; Fox et al., 2005; Golland et al., 2007; Miller & Buschman, 2013; Vincent et al., 2008). The idea of competitive large-scale network interactions emerged when Fox et al. (2005) reported that the DN and “task positive” network were anticorrelated during the resting state, potentially reflecting a toggling between internally-oriented conceptual and externally-oriented perceptual cognitive processing (see also Fransson, 2005). However, subsequent studies led to a refinement of this idea, demonstrating co-activation and positive functional connectivity between the DN and the frontoparietal control network (FPCN)―a component of the “task positive” network―during some task conditions, including mind wandering (Christoff, 2012; Christoff, Gordon, et al., 2009; Fox et al., 2015), spontaneous thought (Ellamil et al., 2016), autobiographical future planning (Gerlach, Spreng, Madore, & Schacter, 2014; Spreng et al., 2010), creativity (Ellamil, Dobson, Beeman, & Christoff, 2012), memory recall (Fornito et al., 2012), working memory guided by information unrelated to current perceptual input (Konishi, 18  McLaren, Engen, & Smallwood, 2015), social working memory (Meyer et al., 2012), and semantic decision making (Krieger-Redwood et al., 2016). Moreover, cooperative dynamics between the FPCN and DN correlate with better task performance in some cases (e.g., Fornito et al., 2012). Cooperative dynamics between these networks may occur when meta-cognitive awareness and/or deliberate control is brought to bear on internally-oriented processing (Andrews-Hanna, Smallwood, et al., 2014; Christoff et al., 2016; Dixon et al., 2014b; Fox & Christoff, 2014; Smallwood, Brown, Baird, & Schooler, 2012). On the other hand, studies have generally found anticorrelation between the DN and other components of the “task positive” network, particularly the dorsal attention network (DAN) (Chai, Castañón, Öngür, & Whitfield-Gabrieli, 2012; Chai, Ofen, Gabrieli, & Whitfield-Gabrieli, 2014; Chang & Glover, 2009; De Havas, Parimal, Soon, & Chee, 2012; Fornito et al., 2012; Gao & Lin, 2012; Holmes et al., 2015; Josipovic, Dinstein, Weber, & Heeger, 2012; Kelly et al., 2008; Lee et al., 2012; Spreng, Stevens, Viviano, & Schacter, 2016; Van Dijk et al., 2010; Yeo, Tandi, & Chee, 2015).  The idea of competitive anticorrelated networks has been influential and used to explain the origin of attentional lapses and behavioral variability in healthy adults (Keller et al., 2015; Kelly et al., 2008; Weissman, Roberts, Visscher, & Woldorff, 2006), cognitive immaturity in children (Chai et al., 2014), and abnormal functioning in conditions such as ADHD (Sonuga-Barke & Castellanos, 2007). Although global signal regression can induce spurious anticorrelations when included as part of data preprocessing (Murphy et al., 2009; Saad et al., 2012), negative FC between the DN and “task-positive” regions has been observed even without this step, suggesting that it is a true biological phenomenon (Chai et al., 2012; Chang & Glover, 2009; Fox, Zhang, Snyder, & Raichle, 2009). However, there are a number of important questions that have not been addressed. We still lack a clear understanding of the strength of negative FC between the DN and DAN; the extent to which the relationship between these networks varies across DN subsystems, different cognitive states, and time; and how DN-DAN interactions relate to broader network dynamics involving the FPCN.  Here, we provide a systematic investigation of DN-DAN interactions to address these questions.  In study 1, we sought to determine the effect size of negative FC between the DN and DAN.  While the notion of anticorrelation is often highlighted in papers that examine DN-DAN interactions, rarely is there discussion of the actual effect size. It is quite possible that negative 19  FC between the DN and DAN is a statistically reliable but weak effect, rather than a true anticorrelation. This is a critical question given that initial studies of anticorrelation used global signal regression (GSR) which is known to alter the distribution of correlation coefficients, and may not provide an accurate assessment of the true effect size of negative FC between the DN and DAN (Murphy et al., 2009). By removing the global signal as part of preprocessing, this mathematically ensures that there are a roughly equal number of positive and negative correlations that are distributed around 0, which can introduce artifactual negative correlations, or inflate the strength of true negative correlations (Murphy et al., 2009). We therefore conducted a meta-analysis of 20 studies reporting anticorrelation to examine empirical effect sizes, and the potential impact of including GSR as part of preprocessing.    In a second study, we examined the variability of DN-DAN interactions in a new data set using several different approaches. Since the discovery of DN-DAN anticorrelation, developments in understanding the DN have now revealed that it is not a unitary entity, but rather, composed of three distinct subsystems (for a review see Andrews-Hanna, Smallwood, et al., 2014). Our first goal was to examine whether the DAN exhibits similar or distinct functional interactions with these subsystems. Although it is too early to definitively characterize the function of each subsystem, preliminary evidence suggests: (1) a Core subsystem involved in self-referential processing, including the construction of a temporally-extended self with attributes, preferences, and autobiographical details; (2) a dorsomedial prefrontal subsystem involved in semantic processing and mentalizing (i.e., generating inferences about mental states including beliefs and desires); and (3) a medial temporal lobe subsystem involved in retrieving and binding together contextual details during the recollection of episodic memories and simulation of future events. Interestingly, studies have found coactivation of the DAN and dorsomedial prefrontal subsystem during a social working memory task (Meyer et al., 2012), and coactivation of the DAN and medial temporal lobe subsystem during a memory-guided attention task (Summerfield et al., 2006), raising the possibility that these subsystems may not be antagonistic with the DAN. Indeed, learning often requires a synergy between perceptual and memory processes (Chun & Turk-Browne, 2007; Hasselmo & McGaughy, 2004), and mental state inferences often draw upon perceptual input (e.g., facial expressions) (Baron-Cohen et al., 2001). Discerning the nature of functional interactions between the DAN and the distinct DN 20  subsystems would provide critical information about the cognitive processes that may or may not be inherently competitive.   A second goal of study 2 was to examine the stability of anticorrelations across time and across different cognitive states. Mounting evidence suggests that the strength and topography of functional connectivity patterns reconfigure across time and different tasks (Allen et al., 2014; Braun et al., 2015; Cole, Reynolds, et al., 2013; Davison et al., 2015; Geerligs, Rubinov, Cam, & Henson, 2015; Gonzalez-Castillo et al., 2015; Hermundstad et al., 2014; Hutchison, Womelsdorf, Gati, et al., 2013; Krienen et al., 2014; Kucyi et al., 2016; Mennes et al., 2013; Shine, Bissett, et al., 2016; Shine, Koyejo, & Poldrack, 2016; Shirer et al., 2012; Simony et al., 2016; Zabelina & Andrews-Hanna, 2016; Zalesky et al., 2014). It is possible that anticorrelations are related to the cognitive state elicited by rest, that is, spontaneous thoughts of current concerns, past events, and future plans (Andrews-Hanna, 2012; Delamillieure et al., 2010). DN-DAN interactions may depart from anticorrelation under some cognitive states, for example, those that require a mixture of perceptual processing and internal conceptual thoughts (Dixon et al., 2014b). A recent study observed DN engagement during an externally-directed working memory task when participants leveraged prior knowledge of the stimuli to complete the task (Spreng et al., 2014), suggesting that there may be task conditions affording greater cooperation between the DN and DAN. Finally, there is some evidence that negative FC involving the DN may vary across time even during rest (Allen et al., 2014; Chang & Glover, 2010). Here, we investigated possible contextual and temporal variability of DN-DAN interactions by examining their relationship across time and different cognitive states within the same participants.   The third goal of study 2 was to examine the possibility that changes across time in DN-DAN FC strength are related to broader temporal dynamics involving the coordination of multiple large-scale networks. Recent work has demonstrated that the strength of FC between a pair of nodes (regions) can increase or decrease across time, and this tends to occur in a coordinated manner, with sets of connections evolving in concert (Bassett et al., 2014; Davison et al., 2015; Zalesky et al., 2014). Here, we sought to extend this idea by examining the temporal co-evolution of interactions at the level of large-scale networks rather than individual regions. Based on evidence that the FPCN has extensive functional interconnections with the DN and DAN (Spreng et al., 2013) and plays a role in regulating internal and external attention (Dixon et 21  al., 2014b; Dixon, Girn, et al., 2017; Gao & Lin, 2012; Smallwood et al., 2012; Spreng et al., 2010; Vincent et al., 2008), we hypothesized that there would be dynamic interactions coordinated across the FPCN, DN, and DAN. For example, we hypothesized that changes across time in the strength of DN-DAN coupling would be tightly coordinated with changes across time in the strength of FPCN-DAN coupling.   To examine these questions, we used fMRI in conjunction with functional connectivity (FC) and machine learning classification analyses. We monitored brain activation dynamics during six conditions: (i) rest; (ii) movie viewing; (iii) analysis of artwork; (iv) social preference shopping task; (v) evaluation-based introspection; and (vi) acceptance-based introspection. Because these conditions differ from traditional cognitive tasks, we refer to them as cognitive states or contexts, rather than tasks. These conditions were designed to elicit mental states that resemble those frequently experienced in everyday life, and were predicted on theoretical grounds to result in variable DN-DAN interactions (Dixon et al., 2014b). That is, we designed conditions that we believed were most likely to show a change in FC away from a negative correlation between the DN and DAN to provide a general test of whether DN-DAN interactions remain stable across different contexts. These conditions involved a combination of internal conceptual and external perceptual processing requirements, or deliberate control over internal processing. Each condition elicited a continuous mental state and did not require any responses. All data underwent the same preprocessing procedure typically used with resting state fMRI that does not rely upon global signal regression (Whitfield-Gabrieli & Nieto-Castanon, 2012). Methods Study 1 effect size meta-analysis In study 1, we examined the effect size of DN-DAN FC in 20 studies. Using Google Scholar and PubMed, we performed searches containing the words: “default network”, “anticorrelation”, “functional connectivity”, and “fMRI”. We found additional studies through the reference lists of these papers. Studies were included in the analysis if they met the following criteria: (i) used fMRI; (ii) acquired data from healthy young adults; (iii) examined DN-DAN FC; and (iv) reported a relevant effect size―an r or z(r) value, or provided figures with legends that allowed 22  for an approximation of the effect size. Because some studies did not report an effect size, our meta-analysis is not exhaustive. Where studies reported results with and without GSR, we included both results for comparison. We report 95% confidence intervals for the median effect size, generated based on bootstrapping with 1000 samples. In most cases studies provided data for a resting state condition, however, there were a few exceptions: Golland et al. (2007) reported data from a movie viewing condition; Fornito et al. (2012) reported “spontaneous” fluctuations reflecting data from a recollection task, after task-related signals had been regressed out; and Amer et al. (2016) reported data from a 1-back task. To calculate the mean effect size across studies, we first Fisher-transformed r-values, averaged them, and then used the inverse Fisher transformation to report the mean r-value. All studies acquired data from healthy adults. However, two studies had unique samples that are worth commenting on. Anderson et al. (2011) used a large sample with ages ranging from 7-35 years (mean = 18.8, SD = 6.1). Although DN-DAN interactions change across development, it is currently unknown when they reach adult-like patterns (Chai et al., 2014; Gao et al., 2013). Thus, it should be kept in mind that the effect size from this study (reflecting data from all participants) may potentially underestimate DN-DAN negative FC. Josipovic et al. (2012) examined DN-DAN FC in a sample of experienced meditators. It is currently unknown whether meditation training leads to enduring changes in resting state network organization, so the finding from this study should be viewed with caution. Additionally, it should be noted that Chang & Glover (2009) only reported data for three participants, and therefore, the values we report from this study likely over-estimate the strength of negative FC. Additional details for each study are presented in Appendix A.  Study 2 participants Participants in study 2 were 24 healthy adults (Mean age = 30.33, SD = 4.80; 10 female; 22 right handed), with no history of head trauma or psychological conditions. This study was approved by the UBC clinical research ethics board, and all participants provided written informed consent, and received payment ($20/hour) for their participation. Due to a technical error, data for the movie and acceptance-based introspection conditions were not collected for one participant. At the end of scanning, another participant reported experiencing physical discomfort throughout the scan. Similar results were obtained with or without inclusion of this participant's data, so they were included in the final analysis. 23  Experimental conditions Each participant performed six conditions in separate six-minute fMRI runs: (1) Rest. Participants lay in the scanner with their eyes closed and were instructed to relax and stay awake, and to allow their thoughts to flow naturally. (2) Movie watching. Participants watched a clip from the movie “Star Wars: Return of the Jedi”, during which Luke Skywalker engages in a light-saber duel with Darth Vader. (3) Artwork analysis. Participants viewed four pieces of pre-selected artwork, each for 90 seconds, and were instructed to attend to the perceptual details and the personal meaning of the art. (4) Shopping task. Participants viewed a pre-recorded video shot from a first-person perspective of items within several stores in a shopping mall, and were instructed to imagine that they were shopping for a birthday gift for a friend, and to think about whether each item would be a suitable gift based on their friend's preferences. (5) Evaluation-based introspection. Participants reflected on a mildly upsetting issue involving a specific person in their life and were asked to analyze why the situation is upsetting, who caused it, what might happen in the future, and to become fully caught up in their thoughts and emotions. (6) Acceptance-based introspection. Participants reflected on a mildly upsetting issue involving a specific person in their life and were asked to cultivate a present-centered awareness, grounded in the acceptance of moment-to-moment viscero-somatic sensations (i.e., to notice and experience arising thoughts, emotions, and bodily sensations with acceptance, and without any elaborative mental analysis or judgment).  Task order was held constant. The introspection conditions were placed at the end so that participants would not continue thinking about the upsetting issue, which may have otherwise influenced thought content during the remaining tasks. Furthermore, because acceptance-based evaluation requires an inhibition of the default tendency to engage in evaluative/narrative processes (Farb et al., 2007), we placed this condition after evaluation-based introspection. Given that the task conditions were completely different and did not require responses, there was no concern about practice effects from one condition to another. That is, there were no specific perceptual or attentional task requirements that participants could improve upon and that could translate from one task condition to another. Additionally, before each of the six conditions we stressed to participants that they should remain as alert as possible, and they reported that they did so (this was confirmed through post-scanning questions regarding attention and the content 24  of each condition). Furthermore, we designed our conditions to be as engaging as possible. Finally, inspection of individual participant data did not reveal evidence of linear changes in DN-DAN FC across the six contexts (Appendix A Figure 1). In fact, Figure 5 reveals that changes in FC across context were brain region specific, and varied in direction (anticorrelations may increase or decrease); no global patterns emerged, suggesting that general factors do not account for our findings.  MRI data acquisition fMRI data were collected using a 3.0-Tesla Philips Intera MRI scanner (Best, Netherlands) with an 8-channel phased array head coil with parallel imaging capability (SENSE). Head motion was minimized using a pillow, and the effect of scanner noise was minimized using earplugs. T2*-weighted functional images were acquired parallel to the anterior commissure/posterior commissure (AC/PC) line using a single shot gradient echo-planar sequence (repetition time, TR = 2 s; TE = 30 ms; flip angle, FA = 90°; field of view, FOV = 240 mm; matrix size = 80 × 80; SENSE factor = 1.0).  Thirty-six interleaved axial slices covering the whole brain were acquired (3-mm thick with 1-mm skip). Each session was six minutes in length, during which 180 functional volumes were acquired. Data collected during the first 4 TRs were discarded to allow for T1 equilibration effects. Before functional imaging, a high resolution T1-weighted structural image was acquired (170 axial slices; TR = 7.7 ms; TE = 3.6 ms; FOV = 256 mm; matrix size = 256 × 256; voxel size = 1 x 1 x 1 mm; FA = 8°). Total scan time was ~ 60 minutes. Head motion was minimized using a pillow, and scanner noise was minimized with earplugs.  Preprocessing Image preprocessing and analysis were conducted with Statistical Parametric Mapping (SPM 8, University College London, London, UK; http://www.fil.ion.ucl.ac.uk/spm/software/spm8).  The time-series data were slice-time corrected (to the middle slice), realigned to the first volume to correct for between-scan motion (using a 6 parameter rigid body transformation), and coregistered with the T1-weighted structural image. The T1 image was bias-corrected and segmented using template (ICBM) tissue probability maps for gray/white matter and CSF.  Parameters obtained from this step were subsequently applied to the functional (re-sampled to 3 25  mm3 voxels) and structural (re-sampled to 1 mm3 voxels) data during normalization to MNI space.  The data were spatially-smoothed using an 8-mm3 full-width at half-maximum Gaussian kernel to reduce the impact of inter-subject variability in brain anatomy.   To address the spurious correlations in resting-state networks caused by head motion, we identified problematic time points during the scan using Artifact Detection Tools (ART, www.nitrc.org/projects/artifact_detect/). Images were specified as outliers according to the following criteria: translational head displacement greater than .5 mm from the previous frame, or rotational displacement greater than .02 radians from the previous frame, or global signal intensity > 4 standard deviations above the mean signal for that session. The mean number of identified outliers was 4.93 (range: 0 - 15) and did not differ across conditions (p > .4). Each participant had at least 5.3 minutes of non-outlier time points. Outlier images were not deleted from the time series, but rather, modeled in the first level general linear model (GLM) in order to keep intact the temporal structure of the data. Each outlier was represented by a single regressor in the GLM, with a 1 for the outlier time point and 0 elsewhere.  Using the CONN software (Whitfield-Gabrieli & Nieto-Castanon, 2012), physiological and other spurious sources of noise were estimated and regressed out using the anatomical CompCor method (Behzadi, Restom, Liau, & Liu, 2007). Global signal regression was not used due to fact that it mathematically introduces negative correlations, and renders the results difficult to interpret (Murphy et al., 2009). The normalized anatomical image for each participant was segmented into white matter (WM), gray matter, and CSF masks using SPM8. To minimize partial voluming with gray matter, the WM and CSF masks were eroded by one voxel. The eroded WM and CSF masks were then used as noise ROIs. Signals from the WM and CSF noise ROIs were extracted from the unsmoothed functional volumes to avoid additional risk of contaminating WM and CSF signals with gray matter signals. The following nuisance variables were regressed out: three principal components of the signals from the WM and CSF noise ROIs; head motion parameters (three rotation and three translation parameters) along with their first-order temporal derivatives; each artifact outlier image; linear trends. A band-pass filter (0.009 Hz < f < 0.10 Hz) was simultaneously applied to the BOLD time series during this step.   26  ROI definition To explore DN-DAN interactions in relation to well-established network boundaries, we used anatomical regions of interest (ROIs) created by Yeo and colleagues (Krienen et al., 2014; Yeo et al., 2015) based on their 17-network parcellation derived from the data of 1,000 participants (Yeo et al., 2011) (Appendix A Figure 2). The 17-network parcellation was split into a set of 114 cortical regions composed of roughly symmetric territories in the left and right hemispheres, and were defined in relation to network boundaries, sulcal patterns, and confidence maps. For each network, spatially connected regions were combined to form a single ROI, whereas spatially disconnected regions became separate ROIs. Vertices near between-network boundaries were peeled back. The current analysis focused on 32 ROIs spanning the DAN and three DN subsystems, and 10 ROIs spanning the FPCN. We extracted the mean activation timeseries from each of these ROIs. Subsystem analysis  To examine whether anticorrelations are present for each DN subsystem, we used the residual timeseries (following nuisance regression) for each ROI to compute condition-specific correlation matrices consisting of all node-to-node connections. After Fisher r-to-z transforming the correlation values, we averaged the z(r) values reflecting pairwise connections between the DAN and each DN subsystem. We first computed average FC separately for the left and right hemispheres, and then averaged them given the similar results; that is, there was no difference between the left and right hemispheres (paired t-tests: ps > .19). This yielded a single value reflecting the relationship between the DAN and each DN subsystem for each participant. These values were submitted to a one-way repeated measures analysis of variance (ANOVA), with subsystem as the factor.   Seed-based voxel analysis We computed seed-based functional connectivity (FC) maps for DAN regions in order to examine the spatial topography of anticorrelated voxels. The timeseries of all voxels within each DAN ROI were averaged, and first-level correlation maps were produced by computing the Pearson correlation between that seed timeseries and the timeseries of all other voxels. 27  Correlation coefficients were converted to normally distributed Fisher transformed z-scores to allow for second-level GLM analyses. Correction for multiple comparisons was accomplished using combined height (Z > 3.1) and cluster (p < .05 FWE corrected) thresholding. Results were visualized with CARET brain mapping software (http://brainmap.wustl.edu/caret; Van Essen, 2005; Van Essen et al., 2001). We examined the location of anticorrelated voxels in relation to the network boundaries from Yeo et al.'s (2011) 17-network parcellation. Similarity analysis To examine potential variability of DN-DAN interactions across the six contexts, we determined the similarity of FC values across contexts. For each participant, we extracted and vectorized all between-network correlations (excluding interhemispheric connections) for each context. Many prior studies have reported stronger within-hemisphere functional connectivity, and it seems likely that interhemispheric functional connections are often indirect, mediated via other brain regions. Thus, we excluded interhemispheric connections to provide more precise results that are likely to reflect direct functional connections. After applying a Fisher’s r-to-z-transform, we used the Pearson correlation as a measure of the similarity of the FC vectors for each pair of contexts. These correlation values were Fisher transformed and averaged, to arrive at a single value reflecting the similarity of FC across contexts. We contrasted across-context similarity with within-context similarity, that is, the similarity of FC values for the early period (first three minutes) and late period (last three minutes) of each condition. The difference between within- and across-context similarity provided an index of the influence of context on DN-DAN FC. Importantly, we computed similarity for each participant separately, and then determined average similarity across the group, thus accounting for individual variability. Machine learning classification analysis We used a support vector machine (SVM) classifier to discern whether an individual's current mental state could be correctly discriminated based solely on DN-DAN FC patterns. Accurate classification would imply a unique configuration of FC values within each context. The SVM classifier was implemented with The Spider toolbox (Weston, Elisseeff, BakIr, & Sinz, 2005). Following prior work (Dosenbach et al., 2010), we set the cost parameter, C, to 1, and used a 28  radial basis function (RBF) kernel, with sigma set to 2 (similar results were obtained with a linear classifier; see Appendix A Figure 3). For each individual we created a vector consisting of all DN-DAN z-transformed correlations (excluding interhemispheric connections) for each context. The correlation vectors served as input features (96 in total), and were assigned a value of 1 or −1 to specify the context to which they belonged. We tested the accuracy of the classifier using leave-one-out cross validation: the classifier was trained on the anticorrelation patterns for all but one participant, and then tested on that left-out participant, and this was repeated for each individual. The methods used for the main analysis were selected a priori. We selected parameters used in prior work (Dosenbach et al., 2010) and did not attempt any type of iterative optimization, and we did not perform any type of feature selection (i.e., all 96 FC values were used). Thus, our analysis method should minimize the chance of overfitting (Skocik et al., 2016). For statistical testing, we obtained an empirical null distribution by performing the classification analysis 1000 times with condition labels randomly permuted. The mean classification accuracy over the 1000 iterations ranged from 49.62% to 50.43% with a standard deviation that ranged from 6.03% to 6.73%, depending on the specific pair of conditions. In each case, inspection of the null distribution revealed that 95% of these models had accuracies below 60.4%. Classification accuracies larger than the 95th percentile of the null distribution were considered to be statistically significant at p < .05. To correct for multiple comparisons, classification accuracies larger than the 99.7th percentile of the null distribution (equivalent to 66.7% accuracy) were considered to be statistically significant at p < .05, bonferroni corrected). To further test the robustness of classification based on DN-DAN FC, we used 4-fold cross-validation in which data were split into 4 equal-sized groups, with 75% of the data used for training the classifier, and the left-out 25% used for testing the classifier. This process was repeated 4 times until every participant was used in the testing set once. In this case, we used a feed-forward neural network classifier that was trained using back propagation using Rapid Miner (Hofmann & Klinkenberg, 2013). The learning rate was set to .3 and momentum was .2. Significant classification was observed with this method as well (Appendix A Figure 4).  Logistic regression analysis We also tested more directly if each condition was associated with distinct FC values by using a logistic mixed-effects modeling approach (Pinheiro & Bates, 2000) implemented with the lme4 29  package in R (Bates, Maechler, & Dai, 2007). This approach allowed us to analyzed the FC values at the item level while modeling the within-subjects variance, as opposed to using a between-subjects approach (e.g., relying on average FC values across participants). Whereas a typical logistic regression (a fixed-effects only model) does not allow for multiple instances per person (violation of the independence assumption), a mixed-effects model deals with non-independence by effectively estimating a random intercept for each individual subject. This ultimately helps to account for the extraneous differences in FC values that are inherently introduced by having multiple observations per subject (Pinheiro & Bates, 2000). The dependent variable was the presence (1) or absence (0) of each condition, yielding six total regressions where each condition was compared against all others. Participant was the random effect in all models, while FC values were fixed effects. A prediction was made for every FC value―was the value from a specific condition (1) or not (0)―while accounting for the within-subjects variance. All significance testing was done using two criteria: (1) a two-tailed α set to 0.05 and (2) a 95% confidence interval as recommended by (Nakagawa & Cuthill, 2007). CIs were determined using bootstrapping with 1000 samples. All six logistic regression models were statistically significant based on these criteria and the models were also significantly different when compared to a random intercept only model (p < .05). Comparisons to the random intercept only model highlight the fact that FC values explained differences in the conditions above and beyond the within-subject variability.   Dynamic FC analysis To examine time-dependent changes in FC during rest, we examined DN-DAN FC within 60 s windows, shifted by one timepoint (2 seconds) each time. Within each window, we calculated the average strength of FC between the DAN and each DN subsystem by computing the mean of the relevant pairwise (node to node) correlations (e.g., averaging the Fisher transformed correlations for each pair of DAN-Core subsystem regions). To limit the possibility of detecting spurious temporal fluctuations in FC, we bandpass filtered the data (0.0167 Hz < f < 0.10 Hz) such that frequencies lower than 1/w were removed, where w is the width of the window (Leonardi & Van De Ville, 2015). We then computed the percentage of windows with z(r) < 0 between the DAN and each DN subsystem, to provide a simple measure of time periods with positive or negative functional coupling. This was done separately for the left and right 30  hemispheres and then averaged given that there was no difference (ps > .23). A one-way repeated measures ANOVA with subsystem as the factor and follow up paired-samples t-tests were used to compare dynamic FC patterns across the DN subsystems.   Temporal co-evolution of network interactions. To examine the temporal co-evolution of interconnected nodes, Bassett and colleagues devised a method of identifying groups of FC connections with statistically similar temporal profiles (Bassett et al., 2014; Davison et al., 2015). This approach first determines the strength of time-varying FC between each pair of nodes (regions), providing numerous time series of edge-weights (connection strength). This approach then uses the correlation coefficient as a measure of the linear association between sets of edge-weight time series, to discern groups of functional connections that display similar changes in strength across time. Here, we adopt this approach, but instead of focusing on node-to-node interactions, we focus on network-to-network interactions defined based on the boundaries of Yeo et al. (2011). Within each 60-second window, we computed the average strength of FC between the DAN and each DN subsystem, and between the FPCN and each DN subsystem, and between the FPCN and DAN. This provided several time-series of between-network FC values reflecting changes across time in the strength of interactions between each pair of networks. We then computed the correlation between each pair of time-series to examine the linear relationship between changes across time in the strength of interactions between each pair of networks. For example, we computed the strength of DN-DAN FC across time and the strength of FPCN-DAN FC across time, and then determined if these changes were correlated. A significant correlation would imply that time-dependent DN-DAN interactions are coordinated with time-dependent FPCN-DAN interactions, and reveal that dynamic changes in FC values across multiple large-scale networks evolve in concert. We computed correlations for the left and right hemispheres separately and then averaged them (following Fisher r-to-z transform) given that the values were highly similar in each case (that is, there was no effect of hemisphere, all ps > .05, corrected for multiple comparisons). To account for the number of correlations performed, we used a Bonferroni correction, such that results at p < .004, uncorrected, were considered statistically significant at p < .05, corrected for multiple comparisons.   We conducted a control analysis to rule out the possibility that temporal co-evolution of network interactions could be driven by participant motion. We examined total motion and 31  framewise displacement. We computed the average amount of motion in each window, just as with between-network FC, and then computed the correlation between changes across time in motion and changes across time in between-network FC for each pair of networks. For each participant we then used the Fisher r-to-z transform of the correlations and determined the mean relationship between temporal variation in motion and between network FC, separately for each type of motion, and each of the six task conditions. These values were submitted to a one-sample t-test to assess statistical significance at the group level, based on α = .05, corrected for multiple comparisons.  Results Effect sizes of functional connectivity between the DN and DAN  Our first question concerned the strength of negative FC between the DN and DAN. To examine this, we summarized effect sizes from 20 studies of DN-DAN interactions (Figure 1; Table 1). We noted a number of variables including whether preprocessing included GSR. As illustrated in Figure 1, studies that used GSR show the expected effect of negative FC between the DN and DAN with a median effect size of r = -.24 (SD = .28; 95% CI: -.50 to -.18). A contrasting picture emerges from studies that did not use GSR. These studies generally show a weak negative correlation or even a small positive correlation between the DN and DAN, with a median effect size of r = -.06 (SD = .20; 95% CI: -.13 to .08). These findings suggest that the DN and DAN may have an independent relationship. Given that GSR is known to shift the distribution of correlation coefficients, this preprocessing step inflates the magnitude of negative FC between the DN and DAN and may give a distorted picture of their associations. Notably, many studies that did not use GSR included multiple preprocessing steps to carefully minimize the effect of noise (e.g., regressing out signals related to respiratory and cardiac effects, white matter and CSF timecourses, and outlier time points) and still reported only weak negative correlations (Table 1).  32   Figure 1. Effect size of DN-DAN functional connectivity across 20 studies. Each point represents mean between-network functional connectivity from one study. Seven studies reported results with and without GSR (global signal regression). Reproduced from Dixon et al. (2017) Table 1. Effect size of DN-DAN functional connectivity across studies Study N Correlation (r) Noise removal Regions Golland et al. 2007 8  0.09 GSR Extrinsic-Intrinsic networks Kelly et al. 2008 26 -0.89 GSR DN-TPN Murphy et al. 2009 12 -0.72 GSR DN-TPN    0.18 3 DN-TPN Chang & Glover 2009 15 -0.35 GSR PCC-DAN   -0.25 2, 3 PCC-DAN Van Dijk et al 2010 98 -0.24 GSR DN-DAN    0.16 1 DN-DAN 33  Study N Correlation (r) Noise removal Regions Anderson et al. 2011 1278  0.05 1, 2 DN-DAN Fornito et al. 2012 16 -0.50 GSR DN-DAN   -0.40 1, 2 DN-DAN Lee et al. 2012 17 -0.74 GSR DN-DAN Chai et al. 2012 15 -0.20 GSR MPFC-DAN   -0.12 1, 2 MPFC-DAN Gao & Lin. 2012 19 -0.20 GSR DN-DAN De Havas et al. 2012 26 -0.26 GSR PCC-DAN Josipovic et al. 2012 14 -0.16 GSR Extrinsic-Intrinsic networks Cole et al. 2014 118 -0.07 GSR DN-DAN Chai et al. 2014 19 -0.15 1, 2 DN-DAN Wotruba et al. 2014 29 -0.21 1, 2 DN-TPN Holmes et al 2015 1570 -0.18 GSR DN-DAN Yeo et al. 2015 68 -0.50 GSR DN-DAN    0.38 1, 2 DN-DAN Spreng et al. 2016 54 -0.04 GSR DN-DAN    0.08 1, 2 DN-DAN Amer et al. (2016) 16 -0.06 1, 2 DN-DAN Current study 24 -0.06 1, 2 DN-DAN 34  Note: Numbers specify preprocessing steps used in studies that did not employ global signal regression (GSR): 1 = regression of motion parameters; 2 = regression of cerebrospinal fluid and white matter timecourses; 3 = regression of respiratory- and cardiac-related signals. DN, default network; DAN, dorsal attention network; TPN, task-positive network; MPFC, medial prefrontal cortex; PCC, posterior cingulate cortex; GSR, global signal regression.   Patterns of functional connectivity between the DAN and each DN subsystem We next examined the variability of DN-DAN interactions. First, we considered regional variability, and examined whether the DAN exhibits similar or distinct patterns of FC with the three DN subsystems during rest. To explore these interactions in relation to well-established network boundaries, we used regions of interest (ROIs) created by Yeo and colleagues (Krienen et al., 2014; Yeo et al., 2015) based on their 17-network parcellation derived from the data of 1,000 participants (Yeo et al., 2011) (Figure 2A; Appendix A Figure 2). We extracted the mean activation timeseries from each of 32 ROIs spanning the DAN and three DN subsystems, and calculated the timeseries correlation between pairs of regions belonging to the DN and DAN. We then computed the average strength of functional connectivity (FC) between the DAN and each DN subsystem. The results demonstrated that DN-DAN interactions significantly varied across DN subsystems [F(2, 46) = 17.78, p < .001] (Figure 2B). The DAN exhibited modest negative FC with the Core subsystem (r = - .13, p < .001), but was uncorrelated with the dorsomedial prefrontal subsystem (r = - .01, p = .56), and showed very weak but reliable negative FC with the medial temporal lobe subsystem (r = - .04, p = .028) (Figure 2B). Negative FC was stronger for the Core subsystem relative to the dorsomedial prefrontal and medial temporal lobe subsystems [t(23) = 5.59, p < .001 and t(23) = 4.02, p = .001, respectively].   Supporting this, whole-brain voxel-wise analyses revealed that DAN seed regions exhibited negative FC with voxels primarily located within the borders of the Core subsystem (Figure 2C; Appendix A Figure 5). Similarly, FC fingerprints for DAN ROIs revealed that negative FC was mainly observed with Core subsystem regions (Figure 2D). Thus, the strength of DN-DAN FC is spatially specific. For example, negative FC was more likely to be observed in the rostromedial prefrontal cortex than adjacent dorsomedial prefrontal cortex. Moreover, 35  region aMT of the DAN did not exhibit anticorrelation with any DN regions. Together, these findings demonstrate regional variability in DN-DAN interactions, with little evidence of negative FC involving the dorsomedial prefrontal and medial temporal lobe subsystems.   Figure 2. Anticorrelation as a function of DN subsystem. (A) Networks from Yeo et al. (2011) used for ROIs. (B) Mean correlation between the DAN and each DN subsystem. Data for each participant (black dots), with mean (red line), 95% CI (red shaded area) and 1 SD (purple lines). (C) Seed-based connectivity analyses showing negative connectivity with DAN regions (Z > 3.1, p < .05 FWE corrected for cluster extent), with the borders of each DN subsystem highlighted. Color bar represents t-values. DAN seeds: FEF, frontal eye fields; aIPS/SPL, anterior 36  intraparietal sulcus/superior parietal lobule; PrCv, ventral precentral cortex; aMT, anterior middle temporal region. Left hemisphere data is presented (see Appendix A Figure 5 for right hemisphere data). (D) Functional connectivity fingerprints for each DAN region. Core subsystem: RMPFC, rostromedial prefrontal cortex; PCC, posterior cingulate cortex; pIPL, posterior inferior parietal lobule; SFS, superior frontal sulcus; rSTS, rostral superior temporal sulcus. DM subsystem: DMPFC, dorsomedial prefrontal cortex, TPJ, temporoparietal junction, TP/LTC, temporopolar cortex/lateral temporal cortex; IFG, inferior frontal gyrus, pDLPFC, posterior dorsolateral prefrontal cortex. MTL subsystem: MTL, medial temporal lobe; RSC, retrosplenial cortex; vpIPL, ventral posterior inferior parietal lobule. Reproduced from Dixon et al. (2017).  Stability of DN-DAN functional connectivity across cognitive states Next, we examined whether DN-DAN interactions exhibit stability across different cognitive states. Prior work has examined the stability of FC patterns by computing the correlation between context-specific connectivity matrices (Cole, Bassett, et al., 2014; Geerligs et al., 2015; Krienen et al., 2014). Strong correlations imply that FC patterns are highly similar across contexts, thus suggesting stability. Here, we adopted this approach, but focused specifically on DN-DAN connections rather than whole-brain FC patterns (Figure 3A). As illustrated in Figure 3B, the similarity between DN-DAN FC patterns across different cognitive contexts was modest. Critically, across-context similarity was significantly lower than within-context similarity―that is, the similarity of DN-DAN FC from the first half to the second half of each context. This was the case when considering all DN as a whole [paired t-test: t(23) = 10.46, p < .001], and when breaking down the analysis by DN subsystem [Core: t(23) = 7.84, p < .001; dorsomedial prefrontal: t(23) = 5.61, p < .001; medial temporal lobe: t(23) = 9.35, p < .001]. The sizable difference between within- and across-context similarity reveals a substantial effect of context on DN-DAN interactions. Importantly, this was not due to the separation of contexts in time; nearly identical results were obtained when comparing FC during one context to FC during the immediately preceding context (Appendix A Figure 6). These findings reveal that DN-DAN 37  interactions vary considerably across different cognitive states. Notably, control analyses ruled out the possibility that the effect of context was driven by motion (see Appendix A Results).  Figure 3. Comparison of within- and across-context similarity of DN-DAN connectivity. (A) Example of the analysis approach for one participant. We extracted DN-DAN correlation values (highlighted by the black box), and then calculated the correlation between the vector of connectivity values for each pair of contexts, and between the vector of connectivity values for the early and late period within each context. (B) Mean within- and across-context similarity of anticorrelations. DAN, dorsal attention network; DN, entire default network; DM, dorsomedial subsystem; MTL, medial temporal lobe subsystem. Error bars reflect within-subject SEM (Loftus & Masson, 1994). Reproduced from Dixon et al. (2017).   We next sought to determine whether it is possible to accurately distinguish the cognitive state of an individual based on a classifier trained only on FC data from other participants. If possible, this would suggest that DN-DAN FC patterns flexibly reconfigure in each context in a manner that is generalizable across participants. A support vector machine (SVM) classifier was fed training data (a vector consisting of all DN-DAN correlations) and learned a model that maximized the separation of two cognitive states (e.g., rest and movie viewing) in multidimensional space, based on the pattern FC values defining each context. The SVM then used its model of the training data to predict the labels of new data. Classifier accuracy was determined using leave-one-out cross validation, and statistical significance was established 38  using permutation testing. As depicted in Figure 4, the SVM achieved classification accuracy that was considerably above chance-level in 12/15 comparisons (ps < .05, uncorrected), and 8 of those comparisons were significant when correcting for multiple comparisons (ps < .05, bonferroni corrected). Supporting the robustness of these results, significant classification was also obtained using 4-fold classification, with 75% of the data used for training and 25% used for testing (Appendix A Figure 4). This suggests that the SVM classifier could distinguish pairs of cognitive states solely on the basis of DN-DAN FC patterns, thereby implying a relatively unique configuration of DN-DAN interactions within each context that was reliable across participants.      Figure 4. Accuracy of the SVM classifier in distinguishing each pair of cognitive contexts. Classification accuracy was significantly above chance level in all cases except for the rest-shopping, shopping-evaluation, and evaluation-acceptance comparisons. Error bars reflect between-subject SEM. *p < .05, uncorrected. **p < .05, bonferroni corrected. Reproduced from Dixon et al. (2017).  To examine more directly whether the six contexts could be discriminated based on FC values we used mixed-effects logistic regression. Rather than pairwise comparisons, this analysis contrasted each context against all others simultaneously. We conducted six regression analyses, and found that each context could be significantly predicted against all others (all ps < .05) [Rest vs others: b = -.33 (95% CI: -.50 to -.16); Movie vs others: b = -1.01 (95% CI: -1.19 to -.82); 39  Artwork vs others: b = -.30 (95% CI: -.47 to -.14); Shopping vs others: b = .28 (95% CI: .09 to .46); evaluation-based introspection vs others: b = .58 (95% CI: .40 to .75); acceptance-based introspection vs others: b = .73 (95% CI: .54 to .92) ]. This provides evidence that each context had a distinct FC pattern from the other five contexts.  We next conducted whole-brain seed-based analyses to provide more detail regarding the direction of changes in DN-DAN FC across different cognitive states. The results demonstrated highly variably patterns (Figure 5). A pair of DN-DAN regions could exhibit negative FC in one context, but no correlation or even positive FC in other contexts (e.g., see aMT-pIPL in Figure 5). Moreover, different region pairs could exhibit changes across contexts in opposite directions. For example, the frontal eye fields and dorsomedial prefrontal cortex exhibited stronger negative FC during the movie condition relative to rest, whereas the anterior intraparietal sulcus and retrosplenial cortex exhibited weaker negative FC during the movie condition relative to rest. These region-specific patterns further underscore regional heterogeneity in DN-DAN interactions. 40   Figure 5. Whole-brain seed-based analyses. Positive and negative functional connectivity for each DAN seed region and context. Negative FC between the DN and DAN flexibly increased and decreased in different cognitive contexts relative to rest. For illustration purposes, we use a slightly liberal threshold to show the full extent of positively and negatively correlated voxels in each context (Z > 2.57, p < .05 FDR cluster corrected). Black star denotes location of DAN seed regions. Right panel: mean FC strength, z(r), for specific pairs of DN-DAN ROIs for each context. Results for the left hemisphere are presented (see Appendix A Figure 7 for right 41  hemisphere data). Based on visual inspection the whole-brain analysis in the left panel, we identified DN regions (indicated with black arrow) that appeared to exhibit an effect of context, and then plotted the mean FC between the DAN and DN seeds for each context in the right panel. This was intended was for illustration purposes only. Color bar shows t-values. Abbreviations: LTC/TP, lateral temporal cortex/temporopolar cortex; DMPFC, dorsomedial prefrontal cortex; RSC/vPCC, retrosplenial cortex/ventral posterior cingulate cortex; pIPL, posterior inferior parietal lobule; RMPFC, rostromedial prefrontal cortex; pDLPFC, posterior dorsolateral prefrontal cortex; IFG, inferior frontal gyrus. Reproduced from Dixon et al. (2017).  Self-reported experience  Contextual modulation of DN-DAN FC may relate to certain aspects of the task conditions we used. Although our conditions did not vary in a systematic manner, we did collect self-reports regarding several variables including the difficulty of the tasks, level of attention, and familiarity and enjoyment with the stimuli in the movie, artwork, and shopping conditions (Table 2). Importantly, participants reported high levels of attention during the conditions with external stimuli. There was a main effect of condition on attention [F(2, 46) = 5.92, p = .005] and enjoyment [F(2, 46) = 40.25, p < .001], with the highest levels in both cases being reported during the artwork condition. The effect of condition on stimulus familiarity was not statistically significant [F(2, 46) = 1.94, p = .16]. Finally, there was a main effect of condition on difficulty [F(4, 92) = 10.97, p < .001], with the movie and artwork conditions being rated as the easiest conditions. Our sample size does not afford enough power for a proper individual differences analysis examining the correlation between these self-report variables and the strength of DN-DAN functional connectivity. However, for completeness we report this information for exploratory purposes in Appendix A Table 2. The fact that participants found the movie and artwork conditions the easiest (and there was the least amount of inter-subject variability) may have resulted in the most distinct FC patterns, and this could potentially explain why the SVM classifier was most accurate in distinguishing these conditions from the others.      42  Table 2. Self-reported experience   Condition Variable Movie Artwork Shopping Evaluation Acceptance Difficulty 1.06 (1.15) 1.73 (1.11) 2.42 (2.06) 3.02 (1.90) 3.88 (1.73) Attention 5.98 (1.03) 6.40 (.071) 5.75 (1.18)   Familiarity/Expertise 3.81 (2.05) 3.90 (1.85) 3.08 (1.67)   Enjoyment of task 4.27 (1.66) 6.38 (.71) 3.00 (1.59)   Note. Participants rated each variable on a 7-point scale from 1=low to 7=high. Values reflect mean across participants with standard deviation in parentheses.    Stability of DN-DAN functional connectivity across time DN-DAN interactions are generally summarized as a single correlation value reflecting connection strength across a long period of time (e.g., 5-10 minutes). While useful, this approach cannot reveal potential temporal variation in DN-DAN interactions. If DN-DAN FC strength is influenced by an individual's current mental state, then it may vary across time even during rest in accordance with changing mental content. We investigated dynamic changes in DN-DAN FC using a 60-second sliding window approach (Hutchison, Womelsdorf, Allen, et al., 2013). Prior work has shown that functionally-relevant FC patterns can be isolated from ~ 60 seconds of data (Gonzalez-Castillo et al., 2015; Leonardi & Van De Ville, 2015; Liegeois et al., 2015; Shirer et al., 2012). For each participant, we computed average DN-DAN FC within each window during rest, and then calculated the percentage of windows during which negative FC was present. The results demonstrated considerable temporal variability, with the DN and DAN alternating between negatively and positively correlated states (Figure 6A). On average, the DAN exhibited negative FC in 67.09% of windows with the Core subsystem, in 52.75% of windows with the dorsomedial prefrontal subsystem, and in 56.16% of windows with medial temporal lobe subsystem (Figure 6B). The number of windows with negative FC varied by subsystem [F(2, 43  46) = 8.95, p =.001], with a higher number for the Core subsystem relative to the dorsomedial prefrontal and medial temporal lobe subsystems (paired t-test: t(23) = 4.38, p < .001 and t(23) = 3.68, p = .001, respectively), recapitulating the distinction between the subsystems observed in the standard analysis. However, even in the case of the Core subsystem there were frequent shifts away from negative FC. Interestingly, temporal variation in FC between the DAN and each DN subsystem followed somewhat unique patterns, highlighting the importance of separating the DN into distinct subsystems rather than treating it as a homogenous network.   Figure 6. Temporal variability in DN-DAN interactions during rest. (A) Data for four randomly chosen example participants demonstrating average correlation strength between the DAN and each DN subsystem within successive 60-second windows. (B) Percentage of windows with negative FC between the DN and DAN. DM, dorsomedial prefrontal subsystem; MTL, medial 44  temporal lobe subsystem. Error bars represent between-subject SEM. Reproduced from Dixon et al. (2017).  Temporal co-evolution of large-scale network interactions  Traditionally, studies have examined temporal variation in the strength of FC between a pair of regions or a pair of networks. However, it is possible that time-varying FC may involve larger coordinated dynamics involving multiple networks. Here, we assessed the possibility that interactions between the DAN and DN evolve across time in a manner that is coordinated with interactions with the frontoparietal control network (FPCN) (Appendix A Figure 8), which has been shown to flexibly couple with these networks. We first computed the strength of FC between each pair of networks within 60-second windows. This provided a time-series of between-network FC values. We then computed pairwise correlations to measure the linear association between the time-series of FC values―our measure of the co-evolution of network interactions. That is, we examined whether sets of between-network connections exhibited statistically similar temporal profiles.   The results demonstrated that periods of time characterized by stronger negative FPCN-DAN coupling were associated with stronger negative DAN-Core coupling (Figures 7A and 7B). This was a robust relationship, observed in every context [all z(r) > .56, p's < .05, Bonferroni corrected]. A similar pattern was observed for the dorsomedial prefrontal (DM) subsystem. In every context, when the FPCN became more negatively coupled with the DAN, the DAN became more negatively coupled with the DM [all z(r) > .54, p's < .05, Bonferroni corrected] (Figures 7A and 7B). A different pattern was observed for the medial temporal lobe (MTL) subsystem of the DN. Changes across time in the strength of FPCN-DAN coupling were unrelated to changes across time in the strength of DAN-MTL coupling (all p's > .05, Bonferroni corrected).  Notably, with one exception, changes across time in the strength of FPCN-DN coupling were unrelated to changes across time in the strength of DAN-DN coupling, and this was true for each of the DN subsystems (all p's > .05, Bonferroni corrected). The one exception was a 45  significant relationship between FPCN-MT coupling and DAN-MT coupling during the movie condition [z(r) = -.21, p < .05, Bonferroni corrected].This suggests that dynamic network co-evolution is specific to particular network interactions and cannot be attributed to a general effect such as global fluctuations in BOLD signal. In particular, when the FPCN became more negatively coupled with the DAN, the DAN became more negatively coupled with the Core and dorsomedial prefrontal subsystems (Figure 8).   Importantly, within each context, temporal variation in the strength of between-network FC was uncorrelated with temporal variation in the amount of participant motion. We found no significant relationships at the group level for total motion [all |z(r)| < .07, p's > .22], or framewise displacement [all |z(r)| < .07, p's > .16]. There was also no evidence of systematic relationships at the level of individual participants. We found 25 out of 284 correlations (~ 9%) were significantly positive at p < .05, bonferroni corrected, and 25 out of 284 correlations (~ 9%) were significantly negative at p < .05, bonferroni corrected. Thus, while some participants did show a significant correlation between temporal variation in the strength of between-network FC and motion in some contexts, this was a rare occurrence, and the correlations were not systematically positive or negative. Thus, temporal co-evolution of network interactions cannot be explained by participant motion.  46   Figure 7. Temporal co-evolution of network interactions. (A) Mean strength of temporal co-evolution. Error bars reflect between-subject SEM. (B) Data for an example participant during the movie viewing condition demonstrating changes across time in functional connectivity between each pair of networks. Top: Changes across time in FPCN-DAN coupling are positively correlated with changes across time in DAN-Core coupling. However, changes across time in FPCN-Core coupling are unrelated to changes across time in DAN-Core coupling. Middle: Changes across time in FPCN-DAN coupling are positively correlated with changes across time 47  in DAN-dorsomedial prefrontal (DM) subsystem coupling. However, changes across time in FPCN-DM coupling are unrelated to changes across time in DAN-DM coupling. Bottom: Changes across time in FPCN-DAN and FPCN-medial temporal lobe (MTL) subsystem coupling are unrelated to changes across time in DAN-MTL coupling. Reproduced from Dixon et al. (2017).  Figure 8. Schematic illustration of temporal co-evolution of network interactions. As the FPCN becomes more anticorrelated with the DAN, the DAN becomes more anticorrelated with the Core and dorsomedial prefrontal subsystems of the DN. As the FPCN becomes more positively correlated with the DAN, the DAN becomes more positively correlated with the Core and dorsomedial prefrontal subsystems. Functional connections between the FPCN and DN Core and dorsomedial prefrontal subsystems are generally positive, and are not shown because they fluctuate across time independently of DN-DAN interactions.  Reproduced from Dixon et al. (2017).  48  Discussion Delineating the nature of functional interactions between the DN and DAN is critical for understanding how attention is efficiently allocated to internal thoughts and external perceptual information. While prior work suggested that the DN-DAN anticorrelation is an intrinsic aspect of functional brain organization based on resting state data, our findings suggest that the DN and DAN have an independent relationship and demonstrate that interactions between these networks exhibit considerable variability: the DAN exhibited differential FC with the three DN subsystems; DN-DAN interactions flexibly reconfigured across different cognitive states; and DN-DAN FC fluctuated across time between periods of anticorrelation and periods of positive correlation. Notably, there was one consistent relationship: temporal fluctuations in FPCN-DAN coupling were correlated with changes across time in the strength of coupling between the DAN and Core and dorsomedial prefrontal subsystems within every context, revealing evidence of temporal co-evolution of large-scale network interactions. Together, these findings suggest that the DN and DAN and the functions they support are not antagonistic, at least in the context of the six different cognitive states that we examined.    Are the DN and DAN anticorrelated? While the notion of anticorrelation is often highlighted in papers that examine DN-DAN interactions, rarely is there discussion of the actual effect size. We therefore conducted a meta-analysis to determine the strength of FC between the DN and DAN, and to examine the influence of global signal regression (GSR) when included as part of preprocessing. Studies that did not use GSR reported weak negative correlations or even positive correlations between the DN and DAN, with a median effect size of r = -.06 (Amer et al., 2016; Anderson, Ferguson, Lopez-Larson, & Yurgelun-Todd, 2011; Chai et al., 2012; Chai et al., 2014; Chang & Glover, 2009; Gao & Lin, 2012; Golland et al., 2007; Murphy et al., 2009; Spreng et al., 2016; Van Dijk et al., 2010; Wotruba et al., 2013; Yeo et al., 2015). The effect sizes suggest more of a weak negative coupling or an independent relationship rather than a competitive anticorrelated relationship highlighting a disconnect between observed effect sizes and the language used to describe DN-DAN interactions. Studies that used GSR found stronger negative FC (median effect size of r = -.24), however, GSR is known to distort the distribution of correlations, making them difficult to 49  interpret (Murphy et al., 2009; Van Dijk et al., 2010). GSR inflates the magnitude of true negative correlations and shifts correlations near r = 0 into artifactual negative correlations.    Several caveats should be taken seriously when interpreting the results of our meta-analysis: (i) we only included studies that reported an effect size and therefore did not perform an exhaustive analysis. Thus, it is quite possible that there are studies showing strong anticorrelation that were not included in this analysis; (ii) there is considerable variability in effect size across studies suggesting that the median effect size reported here should be interpreted cautiously; (iii) studies differed in network definitions; (iv) studies differed in preprocessing steps (aside from inclusion/exclusion of GSR); and (v) there may be some measurement error related to approximating some of the effect sizes from figures. While considering these limitations, this analysis clearly reveals that many studies have observed little to no negative FC between the DN and DAN. This calls into question the idea that these networks are strictly competitive.  It is possible that DN-DAN anticorrelation is a real but transient phenomenon, dependent on cognitive state. The idea that anticorrelation may be a transient rather than persistent aspect of functional network organization is supported by our dynamic FC analysis, which revealed periods of time when the DN and DAN showed strong negative FC (that is, anticorrelation), but also periods of time when these networks exhibited positive FC. Thus anticorrelation may dynamically emerge during some cognitive states, but does not appear to be an invariant feature of functional brain organization. Notably, we also found that region aMT of the DAN exhibited no evidence of negative FC with any DN regions during rest. In fact, during some conditions this region exhibited positive FC with DN regions including the posterior inferior parietal lobule, temporoparietal junction, inferior frontal gyrus, and temporopolar cortex. This finding suggests that aMT may provide a bridge between the DN and DAN, and underscores the fact that these networks are not strictly antagonistic. Together, these results suggest that it is necessary to re-conceptualize the relationship between the DN and DAN, as well as the idea of a competition between internally-oriented conceptual and externally-oriented perceptual processes. Indeed, considerable evidence suggests a more complex picture, with many cognitive states requiring a combination of internally-oriented thoughts and perceptual information (Dixon et al., 2014b).  50   Although the DN and DAN often show differences in overall activation levels in tasks that require perceptual attention versus introspective processing, this does not imply that they must exhibit anticorrelated signal fluctuations. Indeed, evidence suggests that overall activation levels may be orthogonal to functional coupling patterns (Murphy et al., 2016). For example, a recent study found that the posterior cingulate cortex exhibited diminished activation levels during a demanding semantic task, yet simultaneously exhibited increased functional coupling with “task-positive” regions (Krieger-Redwood et al., 2016). Thus, ongoing inter-regional interactions may support information processing that is, to some extent, independent from task-related activation levels. It is important to note that this re-conceptualization of DN-DAN interactions does not take away from the significance of previously reported age-related and group differences in DN-DAN interactions (Chai et al., 2014; Gao et al., 2013; Keller et al., 2015; Spreng et al., 2016). These differences likely contribute to age- and group-related differences in cognitive abilities. The findings reported here have implications for interpreting DN-DAN interactions and the meaning of changes in certain groups of participants, but do not question the differences themselves.    Variable interactions between the DAN and DN subsystems A previous study noted spatial heterogeneity in FC between the DN and DAN, with some connections exhibiting positive correlation and other connections exhibiting negative correlation (Anderson et al., 2011). Here, we extend this work by examining interactions in relation to the well-established division of the DN into three subsystems (Andrews-Hanna, Reidler, Sepulcre, et al., 2010). The DAN exhibited modest negative FC with the Core subsystem, but was uncorrelated with the dorsomedial prefrontal and medial temporal lobe subsystems. These findings are to some extent consistent with Fox et al.'s (2005) original report of DN-DAN anticorrelation, which was based on seed regions located within the Core subsystem, but further emphasize that the DN is not a homogenous network (Andrews-Hanna, Reidler, Sepulcre, et al., 2010; Andrews-Hanna, Saxe, et al., 2014). Even beyond the finding that the DAN exhibited distinct interactions with the three DN subsystems, we found that specific node-to-node connections between the DN and DAN exhibited different patterns of change in correlation strength across contexts. For example, the frontal eye fields and dorsomedial prefrontal cortex exhibited stronger negative FC during the movie condition relative to rest, whereas the anterior 51  intraparietal sulcus and retrosplenial cortex exhibited weaker negative FC during the movie condition relative to rest. Together, these results suggest that a single correlation value reflecting DN-DAN interactions may overlook the variability present at a finer spatial scale, and potentially give a misleading impression of network dynamics.   The DN Core is recruited during a variety of tasks involving self-referential processing (Denny, Kober, Wager, & Ochsner, 2012), value-based decision making (Bartra, McGuire, & Kable, 2013), mind wandering (Fox et al., 2015), autobiographical memory (Andrews-Hanna, Saxe, et al., 2014), and reflection on personal goals (D'Argembeau et al., 2010). This subsystem may therefore play a role in thinking about the self as an object of awareness with particular goals, attributes, and a linear narrative that connects past, present, and future experience―that is, an autobiographical mode of self processing (Araujo, Kaplan, Damasio, & Damasio, 2015; Christoff, Cosmelli, Legrand, & Thompson, 2011; Denny et al., 2012; Farb et al., 2007; Gallagher, 2000; Murray, Schaer, & Debbane, 2012; Schmitz & Johnson, 2007; Wagner, Haxby, & Heatherton, 2012). One possibility is that periods of time characterized by negative FC between the DAN and DN Core subsystem reflects the focusing of attention towards abstract self-related information and away from more concrete perceptual information, whereas periods of positive FC may allow perception to inform self-referential thinking or vice versa. However, there is much still to be learned about the functions of the DN core (e.g., Konishi et al., 2015; Leech, Braga, & Sharp, 2012) and the implications of these dynamics for understanding cognitive functioning.  In agreement with our results, numerous lines of evidence suggest that mentalizing and mnemonic processes that may be associated with the dorsomedial prefrontal and medial temporal lobe subsystems are not inherently antagonistic with perceptual processes associated with the DAN (Dixon et al., 2014b). For example, memory can facilitate the deployment of attention to the external environment (e.g., remembering where one last put the car keys) and this is subserved by co-activation of medial temporal and DAN regions (Summerfield et al., 2006). Similarly, another study found that working memory performance was facilitated for famous relative to unfamiliar faces, and this was accompanied by medial temporal lobe subsystem activation, consistent with the idea that mnemonic representations can facilitate perceptual encoding when it is congruent with task demands (Spreng et al., 2014). Furthermore, during the 52  encoding of new information, medial temporal regions decouple from other DN regions (Huijbers, Pennartz, Cabeza, & Daselaar, 2011), and become more sensitive to afferent sensory input, as a result of acetylcholine's modulatory influence on medial temporal lobe circuit dynamics (Hasselmo & McGaughy, 2004). Finally, during rest, the spontaneous reactivation of information stored in memory may in some cases lead to an autobiographical stream of thought that becomes elaborated upon by the Core subsystem, but in other cases may trigger a sensorimotor stream of thought (e.g., an imagined interaction with the environment) that may elicit cooperative medial temporal lobe subsystem-DAN dynamics. Accordingly, one hypothesis is that the medial temporal lobe subsystem may go in and out of phase with the DAN depending on whether mnemonic and perceptual processes pertain to the same or different goals, thus resulting in uncorrelated activation on average.   Similarly, mentalizing and perceptual processing may sometimes operate in concert, as perception of body language, facial expression, and eye-gaze often inform the inferences we make about others' thoughts, and vice versa (Baron-Cohen et al., 2001). Supporting this idea, coactivation of the DAN and dorsomedial subsystem is observed when individuals view dynamic animations and attend to the social intentional meaning of the movements (Tavares, Lawrence, & Barnard, 2008). Thus, mentalizing and memory processes are sometimes, but not always associated with perceptual decoupling (Schooler et al., 2011; Smallwood et al., 2012). The brain has limited attentional resources, and consequently, has difficulty performing more than one goal at a time (Marois & Ivanoff, 2005). When mentalizing and mnemonic processes can be linked to perceptual processing in service of a unified goal, there may be little to no interference, but when they pertain to different goals (e.g., during task-unrelated thought) they are likely to compete (Dixon et al., 2014b). Alternating anticorrelation and positive correlation between the DAN and these subsystems during rest may reflect the exploration of frequently occurring network states.  Contextual variability of DN-DAN interactions A burgeoning literature has revealed context-dependent FC patterns, with an emerging picture of the brain as a dynamic system that flexibly adapts to changes in internal and external states (Allen et al., 2014; Braun et al., 2015; Cole, Reynolds, et al., 2013; Davison et al., 2015; Fornito et al., 2012; Geerligs et al., 2015; Gonzalez-Castillo et al., 2015; Krienen et al., 2014; Kucyi et 53  al., 2016; Mennes et al., 2013; Milazzo et al., 2014; Shine, Bissett, et al., 2016; Shirer et al., 2012; Simony et al., 2016; Spreng et al., 2010). FC patterns have been linked to individuals' mental states (Andrews-Hanna, Reidler, Huang, et al., 2010; Doucet et al., 2012; Kucyi et al., 2016), and flexibility of FC patterns appears to be adaptive, given that it correlates with task performance (Braun et al., 2015; Fornito et al., 2012; Hermundstad et al., 2014). Building upon this work, we report convergent findings revealing that DN-DAN interactions vary across different cognitive states.   Our similarity analysis revealed little stability in DAN-DN FC across different cognitive contexts. Consistent with this, a prior study found that anticorrelations were more similar from the early period to the late period of a flanker task (r = .61) than between rest and the flanker task (r = .34) (Kelly et al., 2008). This is comparable to the values that we observed, and suggests that DN-DAN interactions are dynamically tailored to one's current context. This complements other work showing context-dependent DN-FPCN interactions (Fornito et al., 2012; Spreng et al., 2010). Furthermore, we found that a machine learning classifier was able to distinguish each pair of contexts solely on the basis of DN-DAN FC patterns. While the classifier's ability to distinguish cognitive states in the current study was noticeably less accurate than results obtained in other studies using whole-brain FC patterns (Gonzalez-Castillo et al., 2015; Milazzo et al., 2014; Shirer et al., 2012), it is quite remarkable that patterns of anticorrelations are sufficiently distinct in each context to allow for above chance-level classification. These findings emphasize flexibility rather than stability in the DN-DAN relationship. Accordingly, DN-DAN interactions during rest do not necessarily reflect the nature of interactions between these networks in general, because other network configurations could occur in other contexts that may be consistent with, or distinct from, the pattern observed during rest. Individual and group differences in anticorrelations during rest could potentially reflect differences in mental state rather than fundamental differences in brain function, although parallel age-related reductions in anticorrelation during task and rest have been observed (Spreng et al., 2016).  Throughout the manuscript we have not emphasized the nature of the task conditions used in the present study because our goal was not to describe the way in which the DN and DAN interact during particular mental states.  Rather, our goal was to test a fundamental hypothesis about the relationship between these networks, and to look for evidence of contextual 54  variability, which we did observe. Our results have broad implications as they robustly demonstrate that DN-DAN interactions are not a stable, fixed feature of brain organization. While our findings suggest a need to re-conceptualize the nature of DN-DAN interactions, our limited range of task conditions means that we cannot specify the principles by which these interactions vary across different cognitive states. Interestingly, we did observe that the SVM classifier was most accurate in distinguishing the conditions that were rated as least difficult and had the least inter-subject variability―the movie and artwork conditions. It is possible that participants were better able to adopt the desired cognitive states in these conditions, providing clear and specific patterns of FC. On the other hand, it is possible that the classifier performed worse at distinguishing the shopping and introspection conditions because participants were less able to consistently adopt the desired cognitive states, resulting in less differentiable FC patterns. Thus, the ease with which participants can perform different instructed tasks may influence the extent to which it is possible to detect reliable variation in FC patterns across contexts. Another possibility is that the movie and artwork conditions were associated with better classification because they were the most structured and stimulus driven and may have constrained FC patterns more than the other conditions that allowed more room for cognitive variability. Temporal co-evolution of large-scale network interactions Network organization dynamically changes across time (Allen et al., 2014; Betzel et al., 2016; Hutchison, Womelsdorf, Gati, et al., 2013; Liegeois et al., 2015; Poldrack et al., 2015; Zalesky et al., 2014), with higher-order association cortices exhibiting considerable flexibility (Braun et al., 2015; Cole, Reynolds, et al., 2013), which may contribute to the context-dependent regulation of thought and perception (Duncan, 2010; Miller & Cohen, 2001). Thus, network neuroscience is now demonstrating a correspondence between the dynamic landscape of network properties and the dynamic nature of cognitive processing. Prior work has shown that sets of functional connections change in strength across time in parallel (Bassett et al., 2014; Davison et al., 2015), and that global brain dynamics exhibit shifts between periods of segregation and integration (Betzel et al., 2016; Liegeois et al., 2015; Shine, Bissett, et al., 2016; Shine, Koyejo, et al., 2016; Zalesky et al., 2014), with between-network connections exhibiting the strongest time-varying dynamics (Zalesky et al., 2014). Although DN-DAN anticorrelation is thought to be a robust feature of brain organization, we observed that DN-DAN interactions alternated across time 55  between periods of anticorrelation and periods of positive correlation. In fact, we found positive FC (r > 0) in about 50% of windows for each of the DN subsystems (slightly fewer windows for the Core subsystem). This suggest frequent transitions between periods of segregation and periods of integration. Prior work offered suggestive evidence that negative FC involving the DN varies across time (Allen et al., 2014; Chang & Glover, 2010). Here, we extend this work by using well-established network boundaries and quantifying the number of windows exhibiting departures from negative FC. Time-dependent interactions between the DN and DAN may provide a balance between functional specialization, and the opportunity for information exchange that allows perception to inform internally-oriented thinking and vice versa.     Using a hypothesis-driven approach, we further found that variation across time in the strength of DN-DAN FC was related to larger patterns of temporal co-evolution between large-scale networks. While prior work has investigated the co-evolution patterns of node-to-node connections across the brain (Bassett et al., 2014; Davison et al., 2015), here we expand on this approach and demonstrate that additional information can be gleaned by constraining such analyses based on theoretical predictions and knowledge of network organization (Yeo et al., 2011). Moreover, our focus on network interactions obviates the need to perform a large number of statistical tests on all time-dependent node-to-node interactions. Within each context, we found that as the FPCN became more anticorrelated with the DAN, the DAN became more anticorrelated with the DN Core and dorsomedial prefrontal subsystems. Interestingly, FPCN interactions with the DN subsystems were not coordinated with DAN-DN interactions suggesting that network co-evolution does not merely represent global changes across the brain, but rather, is spatially specific. It is possible that different network relationships could emerge in other contexts (e.g., greater positive FPCN-DAN coupling may be associated with stronger DAN-Core anticorrelation during a visuospatial working memory task). However, the key point is that our findings provide novel evidence for coordinated changes in FC strength across multiple large-scale networks. Importantly, we found that these temporal changes in between-network FC were uncorrelated with temporal changes in participant motion, suggesting that they are not artifactual.  One possibility is that these structured temporal changes in large-scale network interactions reflect shifting attentional priorities. Abundant evidence suggests that the FPCN 56  encodes task demands, and transmits signals about the current relevance of stimuli, actions, and outcomes to other regions, thus coordinating processing across the cortex (Buschman & Miller, 2007; Cole, Ito, & Braver, 2015; Crowe et al., 2013; Dixon & Christoff, 2012, 2014; Dixon et al., 2014a; Duncan, 2010; Miller & Cohen, 2001; Tomita et al., 1999). Here, we extend these findings by demonstrating that FPCN FC patterns are tightly coupled with the strength of DN-DAN FC changes across time. The large-scale network co-evolution we observed here could potentially reflect moment-to-moment shifts in the distribution of attention between perceptual information and internally-oriented thought. One possibility is that periods of stronger anticorrelation between the FPCN and DAN occurring in concert with stronger anticorrelation between the DAN and the Core and dorsomedial prefrontal subsystems could potentially reflect a state characterized by a decoupling between perceptual processing and abstract thoughts related to self-reflection or mental state inference. Indeed, given that the same relationship was not observed with the medial temporal lobe subsystem, this suggests that the observed network dynamics may relate to the complexity or abstractness of representations, given the roles of the Core and dorsomedial prefrontal subsystems in processing high-level conceptual information related to the self and others (Andrews-Hanna, Saxe, et al., 2014; Binder, Desai, Graves, & Conant, 2009; D'Argembeau et al., 2012; D'Argembeau et al., 2010; Denny et al., 2012; Hassabis et al., 2013; Simony et al., 2016,  but see Konishi et al., 2015). More broadly, examining the temporal co-evolution of network interactions may shed new light on the neural architecture of different cognitive states and how they evolve across time.   Limitations A limitation of the current study is the lack information about the nature and timing of ongoing cognitive activity, and how it relates to variability in DN-DAN interactions. To directly compare FC patterns during various cognitive states and rest, we did not have participants make responses. However, the lack of behavioral data meant that we could not link variation in FC patterns to behavioral performance. Other work has drawn links between task performance and FC patterns (Braun et al., 2015; Cole et al., 2012; Fornito et al., 2012; Kucyi et al., 2016; Schultz & Cole, 2016; Shine, Bissett, et al., 2016), and future studies could further benefit from the use of online experience sampling (Christoff, 2012; Fazelpour & Thompson, 2014) to map the relationship between FC patterns and cognitive states as they evolves across time. Additionally, 57  experimenter controlled variations in task demands on the scale of tens of seconds could also be useful in linking FC patterns to mental states (Gonzalez-Castillo et al., 2015). A second limitation is that we used a limited range of task conditions and cannot specify the principles by which DN-DAN interactions vary across different cognitive states. While we used a hypothesis-driven approach to examine our prediction that DN-DAN interactions are not stable but vary depending on cognitive state, future work could use a range of tasks that systematically vary the required cognitive operations in order to provide additional evidence about the factors that govern contextual variability in DN-DAN interactions. A third limitation pertains to individual variability in network organization. Although we have characterized DN-DAN interactions in relation to well-established network boundaries (Yeo et al., 2011), these boundaries vary across individuals (Mueller et al., 2013). Future work could improve precision by using individually-tailored network ROIs (Wang et al., 2015). Finally, it could be argued that the contextual variation in DN-DAN FC that we observed was due to idiosyncratic numbers of attentional lapses in each context. However, several factors make this very unlikely. First, and foremost, the effect of context was not uniform across all DN-DAN functional connections. For example, from rest to the movie condition, some DN-DAN functional connections exhibited stronger negative FC, while others exhibited weaker negative FC or no change at all. This finding is inconsistent with a general, non-specific factor such as arousal driving the effect of context on anticorrelations. Second, participants reported high levels of attention during the conditions requiring an external focus. Finally, the machine learning classifier was able to accurately discriminate mental states for each participant based on the data from other participants, implying that there was structure in how anticorrelations varied across contexts. Thus, changes in DN-DAN FC across contexts appear to be specifically related to differences in the required cognitive demands.  Conclusions To summarize, we have found that the DN and DAN have a largely independent relationship when GSR is not used as part of preprocessing. Additionally, DN-DAN interactions are more variable than previously appreciated, suggesting that these networks and the functions they support are not strictly competitive. DN-DAN interactions varied across the three DN subsystems, exhibited a high degree of flexibility across different cognitive states, and alternated 58  across time from positive to negative functional coupling. Finally, we found that these changes across time were systematically related to larger patterns of dynamic network co-evolution involving the FPCN, perhaps reflecting shifting attentional priorities. Together, these findings highlight the complexity of interactions between large-scale networks underlying thought and perception.        59  CHAPTER 3 - EXPERIMENT 2: FRACTIONATING THE FRONTOPARIETAL CONTROL NETWORK BASED ON INTER-NETWORK CONNECTIVITY  Introduction The capacity to deliberately regulate attention, often referred to as executive control, is associated with a variety of positive outcomes, including emotional well-being, physical health, academic and financial success, and socially appropriate behavior (Cole, Repovs, et al., 2014; Moffitt et al., 2011). Modern neuroscientific investigations have demonstrated that frontoparietal cortices contribute to executive control via the flexible encoding of task demands and desired outcomes, and the top-down modulation of processing in other brain regions (Braun et al., 2015; Braver, 2012; Cole, Repovs, et al., 2014; Cole, Reynolds, et al., 2013; Dixon & Christoff, 2012; Dosenbach et al., 2006; Duncan, 2010; Miller & Cohen, 2001; Spreng et al., 2010; Stokes et al., 2013). Despite this progress, we still lack a clear understanding of the intrinsic functional organization of frontoparietal cortex, a critical step in discerning the network architecture underlying the deliberate control of attention. Distributed frontoparietal regions often activate together in response to diverse task demands, suggesting that they may function as a unified, domain general control system, referred to as the frontoparietal control network (FPCN) or “multiple demand” network (Duncan, 2010; Vincent et al., 2008). However, it is possible that a finer-level of organization may be present, with distinct subsystems contributing to different types of executive control. Progress has been made in understanding other networks (e.g., default network) via fractionating them into distinct subsystems with unique functional roles (Andrews-Hanna, Smallwood, et al., 2014; Chang, Yarkoni, Khaw, & Sanfey, 2013). Whilst existing models have distinguished the FPCN from networks centered on insular and cingulate cortices (e.g., “salience” and cingulo-opercular networks) (Dosenbach et al., 2007; Seeley et al., 2007), the possibility that the FPCN itself can be fractionated into distinct subsystems with unique functions has yet to be examined.   Several lines of evidence suggest that the FPCN may be composed of  distinct functional territories. First, functional neuroimaging and lesion studies indicate a hierarchical anterior-to-posterior gradient within the lateral prefrontal cortex (PFC), with different sub-regions 60  processing information at different levels of abstraction (Badre & D'Esposito, 2009; Bunge & Zelazo, 2006; Christoff & Gabrieli, 2000; Christoff, Keramatian, et al., 2009; Dixon et al., 2014a; Koechlin, Ody, & Kouneiher, 2003; Petrides, 2005). It is possible that this gradient extends beyond the lateral PFC to large-scale network organization (Dixon, Girn, et al., 2017). Second, anterior frontoparietal regions have expanded during primate evolution (Buckner & Krienen, 2013; Hill et al., 2010; Passingham & Wise, 2012; Teffer & Semendeferi, 2012; Van Essen & Dierker, 2007) and may contribute to uniquely human cognition including higher-order reasoning, advanced mentalizing abilities, and mental time travel (Koechlin, 2011; Suddendorf & Corballis, 2007). Third, developmental patterns suggest a progression from sensorimotor-based cognition to more abstract thinking, commensurate with the protracted maturation of increasingly rostral brain regions and association networks (Buckner & Krienen, 2013; Fair et al., 2007; Teffer & Semendeferi, 2012). Finally, Yeo and colleagues (Yeo et al., 2011) reported that frontoparietal regions fractionate into two networks based on whole-brain functional connectivity (FC) patterns in their 17-network parcellation.   Here, we used a hypothesis-driven approach together with recently developed graph theoretical analyses to reveal a novel organizing principle within frontoparietal cortex. The FPCN is extensively interconnected with both the default network (DN) and dorsal attention network (DAN) (Spreng et al., 2013)―large-scale systems that contribute to distinct, and sometimes competing modes of processing (Dixon, Andrews-Hanna, et al., 2017; Dixon et al., 2014b; Fox et al., 2005). The DAN has a close relationship with sensorimotor regions (Yeo et al., 2011) and plays a key role in visuospatial perceptual attention (Buschman & Kastner, 2015; Corbetta & Shulman, 2002). It contains neurons with spatially organized receptive fields (Moore & Armstrong, 2003) that are activated during saccades (Corbetta et al., 1998), shifts of attention to salient objects in the external environment (Bisley & Goldberg, 2003; Buschman & Miller, 2007; Gottlieb et al., 1998; Ptak, 2012), and during reaching actions towards those objects (Corbetta & Shulman, 2002). In contrast, the DN contributes to more abstract conceptual and associative processes that are, in some cases, independent from sensory input (Andrews-Hanna, Smallwood, et al., 2014; Buckner et al., 2008; Konishi et al., 2015; Margulies et al., 2016). The DN contributes to mentalizing (Spreng et al., 2009), autobiographical memory (Andrews-Hanna, Saxe, et al., 2014), spontaneous cognition (Christoff, Gordon, et al., 2009; Christoff et al., in press; Ellamil et al., 2016; Fox et al., 2015), self-referential processing (Denny et al., 2012), and 61  affective preferences (Andrews-Hanna, Smallwood, et al., 2014). Based on its close relationship with these networks, we hypothesized that the FPCN may be organized into two distinct subsystems related to visuospatial perceptual attention and abstract conceptual thought, respectively. In other words, the distinct processing streams linked to the DN and DAN may be carried forward into the FPCN. While prior work has documented functional connections linking the FPCN to the DN and DAN (Spreng et al., 2013), here we looked for a novel organizing principle within the FPCN by analyzing FC patterns during different cognitive states, examining static and dynamic FC, and examining meta-analytic co-activation patterns across a large number of task conditions.   Given that connectivity patterns constrain function (Passingham, Stephan, & Kotter, 2002), we looked for a potential FPCN fractionation in the functional network architecture. First, we examined whether the FPCN demonstrates evidence of topographically organized functional connections with the DN and DAN. That is, we predicted that FPCN regions coupled with the DN would be spatially distinct from FPCN regions coupled with the DAN. We investigated functional coupling patterns during rest and several different tasks, which allowed us to determine whether differences in coupling patterns persist across different cognitive states. Second, to determine the generalizability of a putative FPCN organization related to the DN and DAN, we examined FC patterns in three independent data sets, and also examined meta-analytic co-activation patterns across 11,406 neuroimaging studies within the Neurosynth database (Yarkoni et al., 2011). Third, we examined the temporal evolution of network interactions, and investigated whether dynamic FC patterns also display evidence of a FPCN fractionation. Specifically, we examined whether spatially-specific FPCN interactions correlate with time-varying changes in the capacity for specialized processing within the DN and DAN, indexed with a graph theoretic measure known as the clustering coefficient (Bullmore & Sporns, 2009; Onnela et al., 2005). Finally, we examined how the putative FPCN fractionation relates to task-related flexibility in FC patterns.  Our primary data set involved data collected from 24 participants that underwent fMRI scanning during six separate conditions that were designed to elicit mental states that are similar to those frequently experienced in everyday life, and that varied in the amount of abstract conceptual thought and perceptual demands: (i) rest; (ii) movie viewing; (iii) analysis of artwork; 62  (iv) social preference shopping task; (v) evaluation-based introspection; and (vi) acceptance-based introspection. Additionally, we examined FC patterns in three other data sets involving traditional cognitive control tasks that are known to activate the FPCN: (i) rule use; (ii) Stroop; (iii) 2-back working memory. Data were processed using a standard approach (Whitfield-Gabrieli & Nieto-Castanon, 2012) without global signal regression in order to avoid distorting FC values (Murphy et al., 2009; Saad et al., 2012).   Methods Participants  Participants in the primary data set were 24 healthy adults (Mean age = 30.33, SD = 4.80; 10 female; 22 right handed), with no history of head trauma or psychological conditions. This study was approved by the UBC clinical research ethics board, and all participants provided written informed consent, and received payment ($20/hour) for their participation. Due to a technical error, data for the movie and acceptance-based introspection conditions were not collected for one participant (S04). At the end of scanning, one participant (S01) reported experiencing physical discomfort throughout the scan. Similar results were obtained with or without inclusion of this participant's data, so they were included in the final analysis. Experimental task conditions The primary data set included six task conditions in separate six-minute fMRI runs. Each task condition was designed to elicit a continuous mental state and did not require any responses. (1) Resting state. Participants lay in the scanner with their eyes closed and were instructed to relax and stay awake, and to allow their thoughts to flow naturally. (2) Movie watching. Participants watched a clip from the movie Star Wars: Return of the Jedi and were instructed to pay attention to the actions of the characters, and also to what they may be thinking and feeling. (3) Artwork analysis. Participants viewed four pieces of artwork in the scanner, each for 90 seconds. These pieces were pre-selected by participants, and during scanning, they were instructed to pay attention to the perceptual details of the art, their inner experience (i.e., thoughts and feelings), and what each image meant to them personally. (4) Shopping task. While in the scanner, participants viewed a pre-recorded video shot from a first-person perspective of items within several stores in a shopping mall. They were told to  imagine themselves going through the mall 63  in order to find a birthday gift for a friend, and to analyze each in terms of whether it would be a suitable birthday gift based on the preferences of their friend. (5) Evaluation-based introspection. Participants were asked to think about a mildly upsetting issue involving a specific person in their life (e.g., a friend, roommate, sibling, or partner), and asked to reflect on what the person and situation means to them, what has happened in the past and may happen in the future, and to analyze everything that is good or bad about the situation. (6) Acceptance-based introspection. Participants were asked to reflect on the same upsetting issue as in the previous case, but this time were instructed to focus on moment-to-moment viscero-somatic sensation, and to accept these sensations without any judgment or elaborative mental analysis.  MRI data acquisition fMRI data were collected using a 3.0-Tesla Philips Intera MRI scanner (Best, Netherlands) with an 8-channel phased array head coil with parallel imaging capability (SENSE).  Head movement was restricted using foam padding around the head.  T2*-weighted functional images were acquired parallel to the anterior commissure/posterior commissure (AC/PC) line using a single shot gradient echo-planar sequence (repetition time, TR = 2 s; TE = 30 ms; flip angle, FA = 90°; field of view, FOV = 240 mm; matrix size = 80 × 80; SENSE factor = 1.0).  Thirty-six interleaved axial slices covering the whole brain were acquired (3-mm thick with 1-mm skip).  Each session was six minutes in length, during which 180 functional volumes were acquired. Data collected during the first 4 TRs were discarded to allow for T1 equilibration effects. Before functional imaging, a high resolution T1-weighted structural image was acquired (170 axial slices; TR = 7.7 ms; TE = 3.6 ms; FOV = 256 mm; matrix size = 256 × 256; voxel size = 1 x 1 x 1 mm; FA = 8°). Total scan time was ~ 60 minutes. Head motion was minimized using a pillow, and scanner noise was minimized with earplugs.  Preprocessing Image preprocessing and analysis were conducted with Statistical Parametric Mapping (SPM8, University College London, London, UK; http://www.fil.ion.ucl.ac.uk/spm/software/spm8). The time-series data were slice-time corrected (to the middle slice), realigned to the first volume to correct for between-scan motion (using a 6 parameter rigid body transformation), and coregistered with the T1-weighted structural image. The T1 image was bias-corrected and 64  segmented using template (ICBM) tissue probability maps for gray/white matter and CSF.  Parameters obtained from this step were subsequently applied to the functional (re-sampled to 3 mm3 voxels) and structural (re-sampled to 1 mm3 voxels) data during normalization to MNI space. The data were spatially-smoothed using an 8-mm3 full-width at half-maximum Gaussian kernel to reduce the impact of inter-subject variability in brain anatomy.    To address the spurious correlations in resting-state networks caused by head motion, we identified problematic time points during the scan using Artifact Detection Tools (ART, www.nitrc.org/projects/artifact_detect/). Images were specified as outliers according to the following criteria: translational head displacement greater than 0.5 mm from the previous frame, or rotational displacement greater than .02 radians from the previous frame, or global signal intensity > 4 standard deviations above the mean signal for that session. The mean number of identified outliers was 4.93 (range: 0 - 15) and did not differ across conditions (F < 1). Each participant had at least 5.3 minutes of non-outlier time points. Outlier images were not deleted from the time series, but rather, modeled in the first level general linear model (GLM) in order to keep intact the temporal structure of the data. Each outlier was represented by a single regressor in the GLM, with a 1 for the outlier time point and 0 elsewhere.   Using the 'CONN' software (Whitfield-Gabrieli & Nieto-Castanon, 2012), physiological and other spurious sources of noise were estimated and regressed out using the anatomical CompCor method (Behzadi et al., 2007). Global signal regression was not used due to fact that it mathematically introduces negative correlations (Murphy et al., 2009). The normalized anatomical image for each participant was segmented into white matter (WM), gray matter, and CSF masks using SPM8. To minimize partial voluming with gray matter, the WM and CSF masks were eroded by one voxel. The eroded WM and CSF masks were then used as noise ROIs. Signals from the WM and CSF noise ROIs were extracted from the unsmoothed functional volumes to avoid additional risk of contaminating WM and CSF signals with gray matter signals. The following nuisance variables were regressed out: three principal components of the signals from the WM and CSF noise ROIs; head motion parameters (three rotation and three translation parameters) along with their first-order temporal derivatives; each artifact outlier image; linear trends. A band-pass filter (0.009 Hz < f < 0.10 Hz) was simultaneously applied to the BOLD time series during this step.  65  Definition of networks and nodes  To analyze network properties, we used anatomical regions of interest (ROIs) created by Yeo and colleagues (Krienen et al., 2014; Yeo et al., 2015) based on their 17-network parcellation derived from the data of 1,000 participants (Yeo et al., 2011). The 17-network parcellation was split into a set of 114 cortical regions composed of roughly symmetric territories in the left and right hemispheres, and were defined in relation to network boundaries, sulcal patterns, and confidence maps. For each network, spatially connected regions were combined to form a single ROI, whereas spatially disconnected regions became separate ROIs. Vertices near between-network boundaries were peeled back. The current analysis focused on 37 ROIs spanning the DN core, DAN, aFPCN, and pFPCN. For the posterior FPCN, we did not use two mid-cingulate ROIs that may extend into white matter and that correlated weakly with other nodes within the network (mean r < .11). For each participant, we extracted the mean timeseries from participants' unsmoothed data for 37 ROIs spanning the DAN, DN, and two FPCNs. Using unsmoothed data minimized the chance of signal contamination across ROIs. The residual timeseries (following nuisance regression) for each ROI was used to compute condition-specific correlation matrices consisting of all node-to-node connections.  Network visualization  For each condition, the group averaged connectivity matrix for each condition was thresholded to retain connections with z(r) > .15, and then submitted to the Kamada–Kawai energy algorithm (Kamada & Kawai, 1989), implemented in Pajek software (De Nooy, Mrvar, & Batagelj, 2011). This algorithm produces spring-embedded layouts that minimize the geometric distances of nodes based on their topological distances in the graph. Well-connected nodes are pulled towards each other, whereas weakly-connected nodes are pushed apart in a manner that minimizes the total energy of the system.  Hierarchical clustering analysis We examined whether FPCN nodes cluster into separate subsystems using two approaches: (i) based on within-FPCN functional connections; and (ii) based on functional connections with the DN and DAN. In each case, the z-transformed correlation matrix was analyzed using hierarchical clustering with the average linkage algorithm (Cluster v3.0, 1988, Stanford University). This 66  approach places strongly correlated nodes within the same cluster, and weakly correlated nodes in separate clusters. Cluster graphs were viewed with Java TreeView (v1.1.6r4 http://jtreeview.sourceforge.net) (Saldanha, 2004). Figure 9 reflects data from left hemisphere nodes, while Appendix B Figure 1 reflects data from right hemisphere nodes.  SVM classification analysis  We used a support vector machine (SVM) classifier to discern whether differences in aFPCN and pFPCN FC patterns were generalizable across participants.  The SVM classifier was implemented with RapidMiner software (Hofmann & Klinkenberg, 2013). The cost parameter, C, was set to 1, and the convergence epsilon was set to .001. Our main analysis used a linear kernel, however, we also report results with an ANOVA kernel to demonstrate the robustness of our results (Appendix B Figure 2). For each individual we created a vector consisting of aFPCN functional connections (z-transformed correlations) to the DN and DAN, and a vector consisting of pFPCN functional connections to the DN and DAN. We excluded interhemispheric correlations which are likely to reflect indirect functional interactions. Additionally, we did not include FC values for one aFPCN region (pre-SMA), so that the aFPCN and pFPCN FC vectors would be equal in length. The correlation vectors served as input features (76 in total), and were assigned a value of 1 or −1 to specify the network to which they belonged. We tested the accuracy of the classifier using 4-fold cross-validation. The data were split into 4 equal-sized groups, with 75% of the data used for training the classifier, and the left-out 25% used for testing the classifier. This process was repeated 4 times until every participant was used in the testing set once. Participants’ data could not appear in both the testing and training set in the same iteration, and we did not perform any type of iterative optimization or feature selection. Thus, our analysis method should minimize the chance of overfitting (Skocik et al., 2016). Notably, when performing the classification analysis 50 times with network labels randomly re-shuffled, mean classification accuracy was at chance level in every condition (Rest: 50.7%; Movie: 51.2%; Artwork: 48.3%; Shopping: 49.4%; Evaluation: 47.4%; Acceptance: 49.6%). Comparing mean between-network FC  After Fisher r-to-z transforming the correlation values, we averaged the z(r) values reflecting pairwise connections between the frontoparietal networks and the DAN and the DN (e.g., the 67  correlations between each pair of aFPCN and DN regions were averaged). We calculated average connectivity separately for the left and right hemispheres, and then collapsed across hemisphere, given the lack of statistical difference (i.e., there was no effect of hemisphere within any condition; all p's > .05). Mean FC values were submitted to a 2 (control network: aFPCN vs pFPCN) x 2 (processing network: DN vs DAN) repeated measures ANOVA.  Replication analyses  We examined whether the FPCN fractionation observed in the main data set would replicate in several independent data sets. In all cases, data were analyzed using the same preprocessing methods as noted earlier. Rule-based cognitive control task. This data set (N = 15) has been described in full elsewhere (Dixon & Christoff, 2012). Participants performed a cognitive control tasks in which they used one of two rules (male/female face discrimination or abstract/concrete word meaning discrimination) to respond to visual stimuli on each trial. On some trials subjects could earn money by responding quickly and accurately. The rules switched from trial to trial requiring participants to actively represent and flexibly switch between the different rules. Data from a single run (run 1 of 6) were analyzed. Stroop Task. This data set (N = 28) was acquired from the OpenfMRI database (accession number ds000164) (Verstynen, 2014). Participants performed the color-word version of the Stroop task with three conditions (congruent, incongruent, and neutral) and were instructed to ignore the meaning of the printed word and respond to the ink color in which the word was printed. Data were acquired in a single run. N-back working memory task. This data set (N = 41) was acquired from the OpenfMRI database (accession number ds000115) (Repovs & Barch, 2012). We analyzed the data from the task period during the 2-back block in control participants. The task was to determine whether each letter was the same as the letter shown two trials previously.    Seed-based voxel analysis  We computed seed-based functional connectivity (FC) maps for aFPCN and pFPCN seed regions in order to examine the spatial topography of positively and negatively correlated voxels with the rest of the brain. The timeseries of all voxels within each ROI were averaged, and first-level correlation maps were produced by computing the Pearson correlation between that seed timeseries and the timeseries of all other voxels. Correlation coefficients were converted to 68  normally distributed Fisher transformed z-scores to allow for second-level GLM analyses. To correct for multiple comparisons, we used a combined height and cluster threshold (Z > 3.1, p < .05 FWE cluster corrected). Results were visualized with CARET brain mapping software (http://brainmap.wustl.edu/caret; Van Essen, 2005; Van Essen et al., 2001). We examined the location of positively and negatively correlated voxels in relation to the DN and DAN network boundaries from Yeo et al.'s (2011) 17-network parcellation. Meta-analytic co-activation maps  We performed a meta-analytic contrast between studies that activated the aFPCN and studies that activated the pFPCN (using network masks from Yeo et al. 2011). The resulting images identify voxels with a greater probability of co-activating with one subsystem or the other. We calculated p values for each voxel using a two-way χ2test between the two sets of studies and thresholded the co-activation images using the False Discovery Rate (q < 0.05). The resulting images were binarized for display purposes and visualized using the NiLearn library for Python. Task-related flexibility  The flexibility index was computed as the similarity of FC patterns within a given condition (from the first half to the second half) minus the similarity of FC patterns between conditions. A larger difference implies that FC patterns changed more across than within conditions, implying task-related flexibility. For each participant and condition, we extracted and vectorized FC values (Fisher transformed correlation values) reflecting aFPCN correlations with DN and DAN nodes, and values reflecting pFPCN correlations with DN and DAN nodes. Prior studies have reported stronger within-hemisphere functional connectivity, and it seems likely that interhemispheric functional connections are often indirect, mediated via other brain regions, and were thus excluded. We used the Pearson correlation as a measure of the similarity of the FC vectors for each pair of conditions. These correlation values were Fisher transformed and averaged, to arrive at a single value reflecting the similarity of FC across conditions for the aFPCN and a single similarity value for the pFPCN. We additionally computed the similarity of FC patterns within each conditions from the first half (first three minutes) to the second half (last three minutes) of each condition. By subtracting between-condition from within-condition similarity values, this provides a selective measure of the effect of condition on FC patterns, with 69  a greater difference indicating stronger task-related flexibility. To determine the flexibility of FC for each FPCN ROI, we computed the variability (SD) of FC across contexts with each DN and DAN node, and then averaged across these values.  Dynamic FC analysis  We conducted a novel dynamic FC analysis to examine the possibility of a FPCN fractionation within the context of dynamic network interactions. Prior work has shown that functionally-relevant connectivity patterns can be isolated from ~ 60 seconds of data (Gonzalez-Castillo et al., 2015; Leonardi & Van De Ville, 2015; Liegeois et al., 2015; Shirer et al., 2012). To examine time-varying connectivity patterns, the data were filtered (0.0167 Hz < f < 0.10 Hz) based on the window size of 60-seconds in order to limit the possibility of detecting spurious temporal fluctuations in FC (Leonardi & Van De Ville, 2015). Within each window, we computed the mean strength of between-network FC for each pair of networks, thus providing time-series of between-network FC values. Within each window we also computed the mean weighted clustering coefficient for the DN and DAN, as a proxy for processing strength. The weighted clustering coefficient reflects the average “intensity” of connections within an interconnected neighborhood. It is computed for each node and usually takes into account all connections to that node (i.e., is agnostic to network boundaries). However, we modified the computation of the weighted clustering coefficient so that only within-network nodes were considered. This served two purposes: (i) it allowed for a meaningful interpretation of resulting clustering values; and (ii) the standard clustering coefficient would have resulted in an artificial correlation between dynamic interactions and changes in clustering strength. For example, because DN nodes are connected with aFPCN nodes, they would normally be included in the computation of the clustering coefficient for each DN node. However, this would create a spurious positive correlation between DN clustering and aFPCN-DN FC strength. Thus, to compute our modified  weighted clustering coefficient, we first extracted relevant within-network connections. We set self-connections and negative FC values to 0, and then normalized all connections by the strongest weight (from the set of all matrices) such that weight magnitudes were rescaled to the range [0,1]. We then took the set of FC values representing within-network connections and computed the weighted clustering coefficient for each node using the Brain Connectivity Toolbox. We then computed the mean weighted clustering coefficient by averaging across 70  clustering values for nodes within a given network (e.g., the DN). The end result was a time-series of mean clustering coefficients for the DN and for the DAN. For each participant, we calculated the correlation between the resulting time-series of between-network FC values and clustering coefficient values (e.g., the time-series characterizing changes in aFPCN-DN coupling was correlated with the time-series characterizing changes in the DN mean weighted clustering coefficient ). We then fisher r-to-z transformed these correlations to allow for statistical testing at the group level. We used a Bonferroni  correction to account for the number of tests performed. Thus, results at p < .005 uncorrected were considered statistically significant at p < .05, Bonferroni corrected for multiple comparisons. Results Graph theory represents complex systems such as the brain as a graph consisting of a set of nodes (regions) and edges (connections between nodes), and allows for a quantitative description of network properties (Bullmore & Sporns, 2009; Fornito et al., 2016; Rubinov & Sporns, 2010). Our nodes were a set of 37 regions of interest (ROIs) created by Yeo and colleagues (Krienen et al., 2014; Yeo et al., 2015) based on their 17-network parcellation (Yeo et al., 2011). The ROIs spanned the DAN, DN (specifically the DN Core subsystem; Andrews-Hanna, Smallwood, et al., 2014), and two frontoparietal networks identified by Yeo et al. (2011) that we refer to as the “anterior subsystem” and “posterior subsystem” of the FPCN (aFPCN and pFPCN, respectively) (Figure 9A). We calculated the time-series correlation between of all pairs of nodes, resulting in a weighted, undirected graph.  Evidence for distinct FPCN subsystems  We first examined whether FPCN nodes separate into distinct subsystems using a hierarchical clustering analysis. Nodes that are strongly connected are placed in the same cluster, while those that are weakly connected are pulled apart and placed in separate clusters. The analysis revealed two distinct clusters corresponding to anterior and posterior FPCN subsystems. These two clusters emerged when considering only within-FPCN functional connections, and when considering “extrinsic” functional connections with the DN and DAN (Figure 9D; Appendix B Figure 1). To examine whether the distinction between aFPCN and pFPCN FC patterns were reliable across participants, we used a linear support vector machine (SVM) classifier to 71  distinguish aFPCN and pFPCN FC patterns in new participants based on data from other participants. The SVM attempts to find a hyper-plane that will best separate the two classes of data. Specifically, we used k-fold cross-validation where the classifier was trained on data from 75% of participants then tested on unlabeled data from the remaining 25% of participants. Using this 4-fold cross validation procedure, we found highly accurate (> 90 %) discrimination of the aFPCN and pFPCN FC patterns during every condition (Figure 10B; Appendix B Figure 2). Permutation testing in which FPCN subsystem labels were randomly shuffled revealed chance level discrimination (~ 50% accuracy; see Methods). These findings provide evidence that the FPCN is not a unitary system, but rather, composed of distinct subsystems.     72  Figure 9. Network ROIs and topology. (A) Control network ROIs based on Yeo et al. (2011). Abbreviations: RLPFC, rostrolateral prefrontal cortex; MFG, middle frontal gyrus; aIPL, anterior inferior parietal lobule; MTG, middle temporal gyrus; pre-SMA, pre-supplementary motor area; aIFS, anterior inferior frontal sulcus; pIFS, posterior inferior frontal sulcus; IPS, intraparietal sulcus; pMTG, posterior middle temporal gyrus; pSFS, posterior superior frontal gyrus; (B) Processing network ROIs based on Yeo et al. (2011). Abbreviations: RMPFC/pgACC, rostromedial prefrontal cortex/pregenual anterior cingulate cortex; PCC, posterior cingulate cortex; pIPL, posterior inferior parietal lobule; SFS, superior frontal suclus; FEFs, frontal eye fields; aIPS/SPL, anterior intraparietal sulcus/superior parietal lobule; PrCv, ventral precentral cortex; aMT anterior middle temporal region. (C) Hierarchical clustering revealed that FPCN nodes group into two separate subsystems. This approach places strongly correlated nodes within the same cluster, and weakly correlated nodes in separate clusters. Hierarchical clustering was performed on FPCN nodes based on within-network connections. (D) Hierarchical clustering was performed on FPCN nodes based on connections with DN and DAN nodes. (E) Support vector machine classifier accuracy in distinguishing aFPCN and pFPCN FC patterns with the DN and DAN during each condition. Dotted line represents baseline accuracy (50%).  Differential coupling patterns with the DN and DAN To elucidate the underlying basis of the FPCN fractionation, we visualized the network topology using the Kamada–Kawai energy algorithm (Kamada & Kawai, 1989), which produces spring-embedded layouts that minimize the geometric distances of nodes based on their topological distances in the graph. Thus, nodes are pulled together or pushed apart based on the strength of functional connections rather than anatomical locations. The network visualization revealed a clear separation of aFPCN and pFPCN nodes during all six conditions, with the former preferentially connecting to DN nodes, and the latter preferentially connecting to DAN nodes (Figure 10). The group-averaged correlation matrix revealed that aFPCN nodes exhibited positive correlations with DN nodes and no correlation or negative correlations with DAN nodes, whereas pFPCN nodes exhibited the opposite pattern (Figure 11A). Furthermore, FC fingerprints (Figure 11B) and whole-brain seed-based correlation maps (Appendix B Figure 3) revealed that spatially adjacent aFPCN and pFPCN nodes exhibited highly divergent functional 73  coupling patterns with DN and DAN regions. Importantly, differences in aFPCN and pFPCN coupling patterns were not driven by spatial proximity to DN and DAN nodes (Appendix B Results).  Figure 10. Visualization of the network topology. Nodes are color-coded based on the 17-network assignment in Yeo et al. (2011). In every context, there is a clear separation between aFPCN and pFPCN nodes, with the former exhibiting preferential FC with DN nodes, and the latter exhibiting preferential FC with DAN nodes.    To quantify these relationships, we computed the average strength of FC between each pair of networks. The aFPCN exhibited strong positive FC with the DN, but was uncorrelated with the DAN, whereas the pFPCN exhibited the opposite pattern (Figure 11). A 2 (FPCN subsystem: aFPCN vs pFPCN) x 2 (processing network: DN vs DAN) repeated measures ANOVA revealed a robust FPCN subsystem x processing network interaction during every condition [rest: F(1, 23) = 105.93, p < .001; movie viewing: F(1, 22) = 152.09, p < .001; artwork analysis: F(1, 23) = 183.10, p < .001; shopping: F(1, 23) = 219.16, p < .001; evaluation-based introspection: F(1, 23) = 98.94, p < .001; acceptance-based introspection: F(1, 22) = 96.83, p < .001]. In each case, aFPCN-DN coupling was stronger than pFPCN-DN coupling (all p's < .05, Bonferroni corrected), whereas pFPCN-DAN coupling was stronger than aFPCN-DAN coupling 74  (all p's < .05, Bonferroni corrected). The DN Core subsystem was our main focus, however, for completeness we also examined whether the dorsomedial prefrontal cortex (DMPFC) and medial temporal lobe (MTL) subsystems of the DN also demonstrated differential FC patterns. The DMPFC subsystem exhibited stronger FC with the aFPCN than the pFPCN in every condition (all p's < .05, Bonferroni corrected), whereas the MTL subsystem exhibited no differences (all p's > .05, Bonferroni corrected) (Appendix B Figure 4).   These findings provide strong evidence that the FPCN fractionation is driven by a preferential relationship between the aFPCN and DN, and a preferential relationship between the pFPCN and DAN. They also suggest that differential coupling may be stronger for the aFPCN than the pFPCN (e.g., see FC fingerprints). To confirm this impression, we computed a “selectivity index” for each node reflecting the strength of preferential FC with the DN versus DAN (see Methods). The mean selectivity index across aFPCN nodes was indeed stronger than the mean selectivity index across pFPCN nodes [paired t-test: t(23) = 4.78, p < .001] (Appendix B Figure 5).   75   Figure 11. Differential FPCN subsystem coupling patterns. (A) Group-averaged correlation matrix reflecting mean z(r) values across the six task conditions. (B) FC fingerprints for each FPCN node (left hemisphere). Top panel: aFPCN nodes demonstrate a clear leftward bias, reflecting stronger FC with DN nodes (yellow text). Bottom panel: pFPCN show a slight rightward bias reflecting stronger FC with DAN nodes (green text), though there is evidence of FC with DN nodes as well. Critically, aFPCN and pFPCN fingerprints are highly divergent for each pair of spatially adjacent nodes (top versus bottom panel fingerprint).    Replication and generalizability of differential coupling patterns 76  To probe the generalizability of the FPCN fractionation, we examined whether it would replicate in three independent data sets involving demanding cognitive control tasks (rule use; Stroop; 2-back working memory). We again found a robust FPCN subsystem x processing network interaction in each data set, with the aFPCN preferentially coupling with the DN, and the pFPCN preferentially coupling with the DAN [rule use task: F(1, 14) = 109.84, p < .001; Stroop task: F(1, 27) = 189.17, p < .001; 2-back working memory task: F(1, 40) = 108.40, p < .001] (Figure 12C).   To examine whether the FPCN fractionation would be present in task-related activation patterns, we performed an automated meta-analysis on coactivation patterns across the wide range of tasks within the Neurosynth database (Yarkoni et al., 2011). The results demonstrated that there are robust differences in co-activation with other parts of the brain between the two networks (Figure 13). In particular, the aFPCN co-activates to a much greater extent with the default network (e.g., rostromedial PFC, posterior cingulate cortex, lateral temporal cortex), than does the pFPCN. There was less evidence for a distinction between the anterior and posterior FPCN with respect to co-activation with the DAN. However, the pFPCN does co-activate to a greater extent portions of DAN adjacent to itself.    77   Figure 12. Mean function connectivity between the FPCN subsystems and the DN and DAN. Conditions are separated into: (A) tasks with a perceptual component; (B) tasks without a perceptual component; and (C) cognitive control tasks from the replication samples. Data for each participant (black dots), with mean (white line), 95% CI (light red and blue shaded areas) and 1 SD (dark red lines).   78  Figure 13. Meta-analytic coactivation contrasts. Red voxels indicate significantly greater coactivation with the pFPCN than aFPCN. Blue voxels indicate significantly greater coactivation with the aFPCN than pFPCN. Images were whole-brain corrected using a false discovery rate of q = 0.05.  Dynamic evolution of differential coupling patterns and network clustering We next asked whether differential coupling is present in dynamic network interactions, and whether these interactions correlate with time-dependent changes in the capacity for specialized processing within the DN and DAN. We quantified the nature of processing within the DN and DAN using a modified version of the clustering coefficient, a graph theoretic measure which computes the number of neighbors around node i that are also interconnected (Rubinov & Sporns, 2010). When normalized by the connection strength (weight) between nodes, the clustering coefficient provides an index of the strength of communication within a densely connected neighborhood (which in this case was a specific network; see Methods). Using a sliding window approach, we derived a time-series of mean weighted clustering coefficients for the DN and DAN, and also derived a time-series of between-network FC values. We then used the Pearson correlation to quantify the relationship between time-dependent coupling patterns and clustering strength.   As illustrated in Figure 14, DN and DAN within-network processing strength (i.e., mean weighted-clustering) varied considerably across time. Critically, these changes were tightly coupled with the strength of interactions involving the FPCN subsystems (Figure 14). Periods of time characterized by stronger aFPCN-DN coupling were associated with larger clustering coefficients for the DN, whereas temporal variation in pFPCN-DN coupling was unrelated to changes in DN clustering. The relationship between aFPCN-DN coupling and DN clustering was significantly stronger than the relationship between pFPCN-DN coupling and DN clustering in every condition (all p's < .05, Bonferroni corrected) except for the acceptance-based introspection condition. On the other hand, periods of time characterized by stronger pFPCN-DAN coupling were associated with larger clustering coefficients for the DAN, whereas temporal variation in aFPCN-DAN coupling was unrelated to changes in DAN clustering. The 79  relationship between pFPCN-DAN coupling and DAN clustering was significantly stronger than the relationship between aFPCN-DAN coupling and DAN clustering in every condition (all p's < .05, Bonferroni corrected) except for the evaluation-based introspection and acceptance-based introspection conditions. Thus, dynamic interactions between the aFPCN and DN are specifically associated with temporal variation in the strength of communication within the DN, whereas dynamic interactions between the pFPCN and DAN are specifically associated with temporal variation in the strength of communication within the DAN.   One potential concern is that participant motion may cause spurious changes in FC. This possibility is unlikely given the specific relationships between dynamic FC patterns and clustering that we observed. However, to rule out this possibility we conducted control analyses and found that motion was uncorrelated with temporal variation in between-network FC and uncorrelated with temporal variation in clustering (Appendix B Results).     Figure 14. Dynamic network interactions and clustering during the Stroop task. (A) Example participant data demonstrating the relationship between temporal fluctuations in FPCN interactions with the DN and DAN and mean weighted clustering strength. (B) Mean correlation 80  between changes across time in clustering and between-network FC strength across participants. Across all conditions, fluctuations in aFPCN-DN FC are positively correlated with fluctuations in DN clustering strength, and fluctuations in pFPCN-DAN FC are positively correlated with fluctuations in DAN clustering strength. Error bars represent between-subject SEM.    FPCN fractionation and task-related flexibility Prior work has shown that FPCN FC patterns exhibit a high-level of task-related flexibility (Cole, Reynolds, et al., 2013; Fornito et al., 2012; Spreng et al., 2010). We examined how the FPCN fractionation relates to flexibility in FC patterns with the DN and DAN. We computed a task-related “flexibility index” reflecting the extent to which FC patterns changed more across conditions than within conditions from the first half to the second half (Methods). Note that this measure of flexibility pertains to context and is different from the measure used by Bassett and colleagues which pertains to flexibility in the temporal domain (Bassett et al., 2011). Both subsystems exhibited a significant flexibility index, revealing task-dependent reconfiguration of FC patterns [aFPCN: t(23) = 5.62, p < .001; pFPCN: t(23) = 8.86, p < .001], (Figure 15A). Interestingly, the pFPCN demonstrated stronger task-related flexibility than the aFPCN [t(23) = 2.25, p = .043]. Not only did overall FC with the DN and DAN change across conditions for both subsystems, but so did the magnitude of the selectivity index―the ratio of DN to DAN connections (Appendix B Figure 6). Thus, while the aFPCN and pFPCN exhibited differential coupling patterns in every condition, the magnitude of this effect was sensitive to task demands. Finally, we found that the right IFS/IFJ node of the pFPCN exhibited the greatest FC variability (Figure 15B; Appendix B 7). In particular, the IFS/IFJ showed stronger FC with DAN regions during the conditions involving a perceptual processing component (movie, artwork analysis, shopping task) than during the conditions that did not involve perceptual processing (rest and introspection conditions), and showed the opposite pattern with DN regions [network x task type interaction: F(1, 23) = 50.26, p < .001] (Figure 15C).    81   Figure 15. Task-related flexibility. (A) Flexibility index reflecting the extent to which FC with the DN and DAN changes across contexts. Both the aFPCN and pFPCN exhibit significant flexibility, but the effect is stronger for the pFPCN. (B) The IFS/IFJ ROI exhibited the greatest FC flexibility across conditions. Mean FC between the IFS/IFJ and each DN and DAN region during perceptual and non-perceptual task conditions. (C) IFS/IFJ seed maps for each condition. For illustration purposes, we use a slightly liberal threshold to show the full extent of positively and negatively correlated voxels (Z > 2.57, p < .05 FDR cluster corrected).   Are the aFPCN and pFPCN subsystems of the same network or extensions of the DN and DAN? An important question is whether the aFPCN and pFPCN should be considered subsystems within the same network (i.e., FPCN) or extensions of the DN and DAN respectively. To examine this, we compared mean between-network FC patterns using paired t-tests and Bonferroni corrected p-values. (Figure 16). During the cognitive control tasks (rule use; Stroop; 2-back working memory), the aFPCN and pFPCN exhibited stronger coupling with each other than with the DN (rule use: p = .057; Stroop: p < .001; 2-back: p = .003) or DAN (rule use: p < .001; Stroop: p < .001; 2-back: p = .008). However, during the other conditions that involved a 82  range of processing demands, the picture is less clear. Coupling between the aFPCN and pFPCN was weaker than aFPCN-DN coupling during the movie and shopping conditions (p's < .05) but not different during the other conditions (p's > .05). Coupling between the aFPCN and pFPCN was stronger than pFPCN-DAN coupling during rest and the evaluation and acceptance introspection conditions (p's < .05) but not different during the other conditions (p's > .05). These findings suggest that the extent to which the aFPCN and pFPCN cluster together versus with the DN/DAN depends on current processing demands.  Figure 16. Mean between-network FC in each condition.    Discussion The FPCN is a central part of the executive control circuitry, enabling individuals to exert deliberate control over attention and flexibly adapt to new situations. The current study provides robust evidence for a novel organizing principle within the FPCN, suggesting that the divergent processing streams supported by the DN and DAN are carried forward into distinct anterior and posterior subsystems of the FPCN: (i) hierarchical clustering revealed a clear separation of aFPCN and pFPCN nodes based on FC with the DN and DAN; (ii) a linear SVM classifier was able to distinguish aFPCN and pFPCN FC patterns with remarkable accuracy; (iii) differential coupling patterns were observed in four independent data sets; (iv) Neurosynth meta-analytic coactivation patterns revealed differential task-based coactivation with the DN and DAN; and (v) dynamic network interactions revealed that the FPCN fractionation is critical for understanding temporal variation in DN and DAN clustering patterns.  83   Brain networks can be understood within the context of a hierarchical gradient of processing. At one extreme unimodal sensorimotor regions process concrete sensory and action-related information, while at the other extreme heteromodal regions elaborate upon such information, allowing for abstract conceptual thought, reasoning, and mental simulations of events (Margulies et al., 2016; Mesulam, 1998; Taylor, Hobbs, Burroni, & Siegelmann, 2015). Recently, it has been shown that the DN occupies a position that is further removed from sensorimotor processing than the DAN based on functional connectivity patterns and spatial distance (Margulies et al., 2016). This is consistent with the idea that the DAN and DN are play different functional roles related to visuospatial perceptual attention and abstract conceptual thought, respectively. Our findings extend this work by demonstrating that these distinct processing streams are carried forward and reflected in the organization of the FPCN.   Functional organization of the FPCN  The DAN is a critical part of the neural circuitry supporting visuospatial perceptual attention. This network is activated when attention is directed in a top-down manner to task-relevant objects and locations, and also when intrinsically salient stimuli are detected (Bisley & Goldberg, 2003; Buschman & Kastner, 2015; Corbetta & Shulman, 2002; Gottlieb et al., 1998; Kastner et al., 1999; Moore & Armstrong, 2003; Ptak, 2012). Our findings suggest a close relationship between the pFPCN and DAN in the network topology. There is compelling evidence that the pFPCN contributes to cognitive control by flexibly encoding task-relevant information including task rules (e.g., stimulus-response mappings) and their relationship to expected reward outcomes (Bunge et al., 2003; Cole, Bagic, et al., 2010; De Baene et al., 2012; Dixon & Christoff, 2012; Duncan, 2010; Miller & Cohen, 2001; Parro et al., 2017; Stokes et al., 2013). Recent work has also demonstrated that pFPCN regions including the inferior frontal junction play a causal top-down role in modulating the DAN and perceptual attention (Baldauf & Desimone, 2014; Bichot et al., 2015; Hampshire et al., 2009). One possibility is that the pFPCN represents information about task context in working memory and that the DAN translates this information into commands to guide the deployment of spatial attention to specific objects and locations (Baldauf & Desimone, 2014; Bichot et al., 2015). By exerting top-down control over the DAN, the pFPCN may ensure that attention remains focused on task-relevant perceptual information, rather than salient, yet irrelevant stimuli, or task-irrelevant thoughts. Thus, the role 84  of the pFPCN in executive control may be related to the abstraction, monitoring, and manipulation of sensorimotor contingencies to facilitate moment-to-moment interactions with the environment.   In contrast, aFPCN regions are rarely recruited during perceptual processing or cognitive tasks that rely on simple rules to respond to stimuli. Rather, the aFPCN is activated when attention is directed towards one’s own thoughts and away from perceptual inputs (Burgess, Dumontheil, & Gilbert, 2007; McCaig et al., 2011), and during high-level cognitive processes that involve monitoring and manipulating abstract information (Badre & D'Esposito, 2009; Christoff & Gabrieli, 2000; Christoff, Keramatian, et al., 2009). Consistent with this, we found that the aFPCN is preferentially coupled with the DN, which plays a role in bringing conceptual/associative knowledge to bear on current thought and perception (Andrews-Hanna, Smallwood, et al., 2014; Konishi et al., 2015; Spreng et al., 2014). The DN has been linked to mental state inference, self-reflection, spontaneous thought, and mental time travel (Andrews-Hanna, Smallwood, et al., 2014; Buckner et al., 2008; Christoff et al., 2016). The notion that the aFPCN has a close relationship with the DN and is involved in internal thought and abstract processing provides a unifying thread that links aFPCN activation across a range of domains including meta-cognitive awareness (Baird, Smallwood, Gorgolewski, & Margulies, 2013; De Martino, Fleming, Garrett, & Dolan, 2013; Fleming et al., 2010; McCaig et al., 2011; McCurdy et al., 2013), relational reasoning (Christoff et al., 2001), multi-tasking and future-directed control signals (Braver & Bongiolatti, 2002; Koechlin et al., 1999 ; Nee & D'Esposito, 2016), retrieving complex task sets from long-term memory (Cole, Bagic, et al., 2010; Sakai, 2008), stimulus-independent and abstract thinking (Burgess et al., 2007; Christoff, Keramatian, et al., 2009; Christoff, Ream, Geddes, & Gabrieli, 2003; Dixon, Girn, et al., 2017; Fox et al., 2015; Smallwood et al., 2012), strategic mentalizing (Bhatt, Lohrenz, Camerer, & Montague, 2010), episodic memory (Gilbert et al., 2006), future planning (Spreng et al., 2010), prospective memory (Poppenk et al., 2010), and future-oriented reward processing (Dixon et al., 2014a; Dixon, Girn, et al., 2017; Jimura, Chushak, & Braver, 2013). Moreover, numerous studies have now found coactivation of FPCN and DN regions, particularly when executive control processes direct internal mentation (Addis, Wong, & Schacter, 2007; Christoff, Gordon, et al., 2009; Christoff et al., 2016; Fornito et al., 2012; Gerlach et al., 2014; Smallwood et al., 2012; Spreng et al., 2010). For example, Fornito et al. (2012) found that FPCN-DN coupling increased during 85  memory recollection and was predictive of efficient behavioral performance. The aFPCN has been linked to high-level mnemonic control processes (Simons & Spiers, 2003) and may allow abstract conceptual information (e.g., one's intentions) to deliberately trigger the mental (re)construction of a memory or imagined future event (Addis et al., 2007; Dixon et al., 2014b; Poppenk et al., 2010; Spreng et al., 2010). More broadly, the aFPCN may contribute to a diverse range of tasks by directing attention to abstract internal goals and conceptual information, thus enabling modes of thought that are relatively free from the constraints of concrete sensorimotor interactions with the environment. A recent framework (Christoff et al., 2016) suggests that aFPCN regions (in particular the rostrolateral prefrontal cortex), may contribute to the abstract “top-level management” of thought, exerting a general constraint that keeps one's focus on task-relevant material, yet allowing for some degree of spontaneous variability in thought. This regulation of internal mentation may facilitate mental time travel, reasoning, planning, creativity, and problem solving. It may also support the performance of traditional cognitive control tasks by allowing abstract task representations to guide the implementation of more concrete rules and actions (Badre & D'Esposito, 2009; Cole, Laurent, & Stocco, 2013; Koechlin & Summerfield, 2007).   In every condition, including demanding cognitive control tasks (rule use, Stroop, 2-back), we found robust coupling between the aFPCN and DN. Consistent with this, a recent study found encoding of task-relevant information by the DN and increased activation during demanding rule switches, suggesting that it may contribute to some forms of cognitive control that involve activating different cognitive contexts (Crittenden et al., 2015). We did find, however, that the magnitude of aFPCN-DN coupling was reduced during the cognitive control tasks relative to other conditions, and was significantly lower than aFPCN-pFPCN coupling. The aFPCN was strongly aligned with the DN across all of the task conditions that were designed to elicit mental states that resemble those frequently experienced in everyday life (the six conditions in the primary data set). Thus, the diminished relationship with the DN during the traditional cognitive control tasks may represent the exception rather than the rule. The aFPCN may typically operate as an extension of the DN, but become co-opted by the pFPCN when it is necessary to perform complex perceptually-focused tasks. Thus, while the pFPCN may have evolved as an extension of the DAN processing stream to allow for the regulation of visuospatial perception based on current task demands, the aFPCN may have evolved as an extension of the 86  DN processing stream to allow for the regulation of abstract thought, including complex social reasoning, future goal planning, and self-reflection. This proposal aligns with suggestion that there is an intimate relationship between brain evolution, including expansion of the anterior prefrontal cortex in humans (Teffer & Semendeferi, 2012), and the emergence of complex social life (Dunbar & Shultz, 2007).     Dynamic FC and clustering While static FC has been the primary focus of past work, studies are increasingly examining the dynamic nature of FC patterns (Bassett et al., 2014; Dixon, Andrews-Hanna, et al., 2017; Hutchison, Womelsdorf, Allen, et al., 2013; Shine, Bissett, et al., 2016; Zalesky et al., 2014). The veracity and functional relevance of dynamic FC has been questioned (Laumann et al., 2016), however, numerous studies have found systematic relationships in how network properties evolve across time (Allen et al., 2014; Davison et al., 2015; Dixon, Andrews-Hanna, et al., 2017; Hutchison, Womelsdorf, Gati, et al., 2013; Zalesky et al., 2014), and have observed correlations with behavior (Braun et al., 2015; Shine, Bissett, et al., 2016; Shine, Koyejo, et al., 2016; Simony et al., 2016) that are independent of extraneous factors such as motion. Here, we tracked the dynamic evolution of network properties and observed a novel relationship between FPCN FC patterns and the strength of clustering within the DN and DAN. The clustering coefficient characterizes the fraction of a node’s neighbors that are also interconnected and may provide a window into the strength of communication within a given network (Rubinov & Sporns, 2010). We found highly specific relationships: changes across time in DN clustering were correlated with aFPCN-DN coupling strength, but not pFPCN-DN coupling strength, and changes across time in DAN clustering were correlated with pFPCN-DAN coupling strength, but not aFPCN-DAN coupling strength. These highly specific interactions were unrelated to motion, suggesting that dynamic FC patterns are functionally significant, and associated with temporal variation in the nature of information processing within specialized processing networks. Our findings are in alignment with electrophysiological recordings in non-human primates demonstrating that control regions transiently modulate processing within specialized neural populations based on task requirements (Bichot et al., 2015; Miller & Buschman, 2013). However, the findings reported here do not speak to directionality and whether clustering was influenced by between-network interactions, or vice versa. The spatially-specific dynamic FC 87  patterns observed here complement prior human FC studies that have documented shifts across time in global network topology from more integrated to more segregated states (Betzel et al., 2016; Liegeois et al., 2015; Shine, Bissett, et al., 2016; Shine, Koyejo, et al., 2016; Zalesky et al., 2014). Future work will be necessary to delineate the functional/behavioral consequences of coordinated changes across time in network interactions and clustering.  Relation to other models of executive control and frontoparietal organization Several large-scale networks contribute to executive control. According to one model, the FPCN is critical for trial-by-trial adjustments in control, whereas a cingulo-opercular network is critical for the maintenance of task-goals across trials, supporting a balance between flexibility and stability (Dosenbach et al., 2006). Rapid adjustments in control may occur via flexible task-dependent shifts in FPCN coupling patterns (Braun et al., 2015; Cole, Reynolds, et al., 2013; Dwyer et al., 2014; Fornito et al., 2012; Spreng et al., 2010) that facilitate the integration of information across distributed systems (Cocchi, Zalesky, Fornito, & Mattingley, 2013). Another model suggests that the “salience” network initiates shifts in modes of information processing related to the FPCN and DN (Sridharan et al., 2008). Our findings suggest an orthogonal dimension of executive control, with different zones within the FPCN involved in visuospatial perceptual attention and more abstract conceptual thought processes, respectively. This division of labor between posterior and anterior parts of the FPCN is compatible with the idea of a rostro-caudal gradient in the lateral PFC based on abstractness of representational content (Badre & D'Esposito, 2009; Christoff & Gabrieli, 2000; Christoff, Keramatian, et al., 2009; Dixon et al., 2014a; Koechlin et al., 2003; Nee & D'Esposito, 2016; O'Reilly et al., 2010). Here, we extend this idea to the large-scale network organization within which the lateral PFC is embedded (Dixon, Girn, et al., 2017). Recent work suggests a “distance from sensory-motor processing” organizational principle, with more complex and abstract processing occurring in regions that are physically remote from primary sensory and motor cortices (Margulies et al., 2016; Taylor et al., 2015). Specifically, frontoparietal and default network regions are located furthest from primary sensory cortices based on connectivity patterns and spatial distance and are preferentially associated with more abstract functions such as reasoning and imagination (Margulies et al., 2016; Taylor et al., 2015). In line with this, Buckner and Krienen (Buckner & Krienen, 2013) suggest that association cortex acquires high-level functions during the course of development 88  because it is spatially distant from primary sensory-motor cortices and not subject to the constraints imposed by extrinsic sensory signals. Notably, our findings suggest that the aFPCN may be further removed from sensory-motor processing than the pFPCN. Consistent with this, we observed that aFPCN but not pFPCN nodes were negatively correlated with primary sensory-motor regions (Appendix B Figure 3). Furthermore, a forward inference meta-analysis revealed that the pFPCN but not the aFPCN is consistently recruited in studies that use the terms “perception” or “action” (Appendix B Figure 8). Thus, a general principle of functional organization may apply across different brain networks (Margulies et al., 2016) and within the FPCN itself.   Other work has emphasized that the FPCN is a flexible hub that coordinates processing across other networks in a task-dependent manner (Braun et al., 2015; Cocchi et al., 2013; Cole, Reynolds, et al., 2013; Spreng et al., 2010; Zabelina & Andrews-Hanna, 2016). In the current study we found that although the aFPCN and pFPCN were aligned with the DN and DAN, respectively, there was evidence that FC patterns flexibly adapted to task demands. There were overall shifts in aFPCN and pFPCN coupling patterns, and also shifts in the relative “preference” of coupling with the DN or DAN. Thus, while there is a broad organizational principle that distinguishes the aFPCN and pFPCN, this reflects a relative and flexible difference in FC patterns rather than an absolute and fixed aspect of network architecture. The organization noted here is thus fully compatible with findings of task-dependent reconfiguration of FPCN FC patterns. However, an important question for future work to address is the extent to which flexible FPCN patterns reflect overall changes in FC with other networks versus changes in the relative strength of FC between the aFPCN and pFPCN and their “preferred” processing networks. Interestingly, we observed a weaker “selectivity index” for the pFPCN, yet greater task-related flexibility across the task conditions examined here. This suggests that pFPCN may have access to information from both the DN and DAN, and may be positioned to flexibly mediate interactions between more concrete visuospatial and more abstract conceptual information. In particular, we found that the IFS/IFJ of the pFPCN exhibited the strongest task-related variability in FC patterns. Prior work has demonstrated a role for this region in encoding task demands (Brass, Derrfuss, Forstmann, & von Cramon, 2005; Dixon & Christoff, 2012) and top-down control of attention (Bichot et al., 2015) via shifting coupling patterns with different regions based on the target of attention (Baldauf & Desimone, 2014). This region may thus play 89  an important role in coordinating DN and DAN processing based on current goals. It could be the case that task-related flexibility was lower for the aFPCN because it is exclusively interconnected with multimodal regions that process highly abstract information, and may not be equipped to interact with many regions. Additionally, it remains possible that a different battery of tasks other than those used here would reveal greater FC flexibility in the aFPCN than the pFPCN.      Limitations There are several limitations to the current study that should be noted. Given that our range of tasks was not exhaustive, it remains possible that other FPCN organizations could potentially emerge in other contexts. Indeed, the functional distinction between the aFPCN and pFPCN likely depends on task demands. In some contexts, such as demanding tasks that require considerable effort, the FPCN may not fractionate, but rather, function as a domain general resource to support task performance (Duncan, 2010). Additionally, there may be some tasks that elicit cooperation between perceptual and abstract conceptual information that may be associated with distinct network interactions (e.g., positive coupling between the aFPCN and DAN). One example may be perceptual metacognition. Part of the aFPCN―the rostrolateral prefrontal cortex―plays a causal role in accurately representing one's knowledge during perceptual decision making (Fleming et al., 2010). Second, future work incorporating behavioral indices and subjective reports regarding perception and abstract thought could provide corroborating evidence about the functional distinctions between the aFPCN and pFPCN. Third, while considerable evidence suggests that frontoparietal cortices exert top-down influences on other regions/networks, the present data do not allow us to draw such conclusions. Moreover, we didn't directly examine executive control in relation to thought and perception. Finally, our analysis is limited by the reliance on pre-defined network boundaries and the assumption of discrete brain clusters/networks. The Yeo et al. (2011) parcellation is a dimensionality reduction on a complex space, and should be viewed as a general guiding principle rather than a set of fixed and precise brain network demarcations. Moreover, the network affiliation of a given brain region can shift across time and context (Bassett et al., 2011; Pessoa, 2014). That being said, our results provide evidence that spatially distinct parts of the FPCN―as defined here―are differentially coupled with the DN and DAN across a range of contexts.  90  Conclusions  Explicating the nature of large-scale network interactions is a critical step towards a systems-level description of how executive control is implemented. Given the multifaceted nature of executive control processes, it seems likely that they rely on multiple interacting, yet distinct neural systems (Reineberg et al., 2015). The current work provides a significant step in discerning the network basis of executive control by revealing a novel organizing principle within frontoparietal cortex. This may offer new predictions about clinical deficits in control functions. For example, altered connectivity between the aFPCN and DN may interfere with regulating abstract information (e.g., self-referential thoughts) in conditions such as depression, whereas altered connectivity between the pFPCN and DAN may interfere with regulating visuospatial attention (e.g., focusing on goal-relevant objects) in conditions such as ADHD. More broadly, our results are consistent with the notion that brain network organziation may, in part, reflect a gradient of representational abstraction.     91  CHAPTER 4 - EXPERIMENT 3: DISTINCT BRAIN NETWORK CONFIGURATIONS DURING EVALUATION-BASED AND ACCEPTANCE-BASED INTROSPECTION  Introduction Humans spend a considerable amount of time engaged in introspective processes, during which attention is directed internally to thoughts and feelings related to memories, future goals, imagined social interactions, daydreams, work demands, and other concerns. Contemplative traditions, modern neuroscience, and clinical psychology have emphasized that attention can be directed to different aspects of subjective experience in ways that promote or hinder well-being (Farb et al., 2007; Gross, 1998; Hayes et al., 2006; Kabat-Zinn, 2009; Segal et al., 2013). Evaluation-based introspection refers to the process of analyzing the meaning of thoughts and emotions within a personal narrative extending from past to future, often accompanied by the desire to experience positive feelings and avoid uncomfortable feelings. While common, this type of introspection can sometimes lead to maladaptive ruminative patterns that exacerbate  negative affect (Nolen-Hoeksema, 2000). In contrast, acceptance-based introspection refers to the non-judgmental observation of thoughts and emotions as passing mental events. This type of introspection is a core aspect of mindfulness practice, and has the aim of cultivating a present-centered awareness grounded in the acceptance of moment-to-moment viscerosomatic sensations, whether pleasant or unpleasant. This type of introspection is thought to promote adaptive emotional responses by minimizing cognitive reactivity, encouraging free-flowing mental activity, and enhancing one's tolerance of uncomfortable feelings (Arch & Craske, 2006; Farb et al., 2010; Segal et al., 2013). In contrast to other forms of emotion regulation, negative emotions are not seen as problematic and requiring change; rather, they are viewed as transient bodily sensations that naturally come and go.  A growing number of neuroscientific studies have examined the relationship between patterns of brain activity and various forms of internally-oriented attention (Andrews-Hanna, Reidler, Huang, et al., 2010; Andrews-Hanna, Smallwood, et al., 2014; Christoff, Gordon, et al., 2009; Ellamil et al., 2016; Farb et al., 2007; Gorgolewski et al., 2014; Konishi et al., 2015; Smallwood et al., 2012; Tusche, Smallwood, Bernhardt, & Singer, 2014). A seminal study by 92  Farb and colleagues (2007) had participants reflect on positive and negative trait adjectives (e.g., melancholy, confident) and adopt an evaluative/narrative mode of introspection during which they thought about the meaning of the word with reference to their temporally extended self-narrative, or adopt an experiential/acceptance mode of introspection during which they simply monitored their present moment experience elicited by the words. The results demonstrated that narrative evaluation was associated with default network (DN) activation, whereas experiential acceptance was associated with activation of the frontoparietal control network (FPCN) and salience network (SN) in meditators relative to novices (Farb et al., 2007). Consistent with these results, the DN is reliably engaged when individuals are asked to make judgments about their own characteristics (Denny et al., 2012), think about what is good or bad for the self (Dixon, Thiruchselvam, Todd, & Christoff, 2017), or imagine events from the past and future (Andrews-Hanna, Smallwood, et al., 2014; Buckner et al., 2008). The DN may thus play a role in conceptualizing and evaluating mental content in relation to knowledge and feelings about “me”. In contrast, the FPCN and SN have been associated with meta-cognitive awareness (Fleming et al., 2010; McCaig et al., 2011), attentional regulation (Duncan, 2013), and interoceptive processing (Craig, 2002; Critchley & Harrison, 2013; Farb et al., 2012; Seeley et al., 2007). These networks may provide a more objective and embodied representation of inner thoughts and feelings as they exist in the present moment, in contrast to the DN which filters experience through a model of desired self-image and circumstances.   Beyond examining brain activation magnitude, critical insights can be gleaned by delineating large-scale network properties, which provide a window into the global information processing landscape underlying complex cognitive processes (Fornito et al., 2016; Medaglia, Lynall, & Bassett, 2015; Petersen & Sporns, 2015; Rubinov & Sporns, 2010). This may be especially important for discerning the neurobiology of introspection, which likely relies on effective communication between distributed cortical and subcortical systems. There is growing evidence for time- and state-dependent changes in functional network architecture (Dixon, Andrews-Hanna, et al., 2017; Fornito et al., 2012; Hutchison, Womelsdorf, Allen, et al., 2013; Krienen et al., 2014; Zalesky et al., 2014), which potentially suggests that network states dynamically reorganize in line with current cognitive operations. Supporting this idea, recent studies have found correlations between indices of network reorganization and behavioral performance on cognitive tasks (Braun et al., 2015; Dwyer et al., 2014; Fornito et al., 2012; Shine, Bissett, et al., 93  2016). However, little is known about the relationship between network organization and different subjective experiences during internally-oriented attention. This is a critical gap given that humans spend considerable time in task-free introspective states that are intimately linked to emotional well-being. While numerous studies have compared network properties in meditators versus non-meditators during rest or a meditative state (Brewer et al., 2011; Farb et al., 2007; Hasenkamp & Barsalou, 2012; Jang et al., 2011; Jao et al., 2016; Josipovic, Dinstein, Weber, & Heeger, 2011; Pagnoni, 2012; Xue, Tang, & Posner, 2011), there have been few studies that have directly contrasted functional network organziation in the same individuals during different types of introspection.  Here, we employed graph theoretic tools to delineate the large-scale network organization during evaluation-based and acceptance-based introspection in participants that were not pre-selected for meditation experience. Graph theory offers tools for quantifying complex systems by modeling networks as nodes (regions) and edges (connections). This approach allows the relationships between multiple nodes to be simultaneously considered and summarized along various dimensions. We compared the two modes of introspection along a number of topological properties. Furthermore, to directly link network properties to subjective experience, we contrasted network organization in “high acceptors” and “low acceptors” based on subjective reports obtained following the scanning sessions. Our results demonstrate that distinct styles of introspection are associated with dynamic reorganization of large-scale network properties. Methods Participants Participants were 24 healthy adults (Mean age = 30.33, SD = 4.80; 10 female; 22 right handed), with no history of head trauma or psychological conditions. This study was approved by the UBC clinical research ethics board, and all participants provided written informed consent, and received payment ($20/hour) for their participation. Due to a technical error, data for the acceptance-based introspection condition was not collected for one participant, resulting in a final sample of 23 participants. At the end of scanning, one participant reported experiencing physical discomfort throughout the scan. Similar results were obtained with or without inclusion of this participant's data, so they were included in the final analysis. 94  Task conditions Prior to scanning, participants underwent a brief training procedure to ensure that they understood the distinct forms of introspection. Then in the scanner, they were asked to adopt each mode of introspection during separate 6-minute scans. Instructions were repeated just prior to each scan and delivered by an author (MLD) with over 16 years of meditation experience.  Resting state. Prior to the evaluation and acceptance conditions, there was a resting state scan during which participants lay in the scanner with their eyes closed and were instructed to relax and stay awake, and to allow their thoughts to flow naturally. Evaluation-based introspection. Participants were asked to think about a mildly upsetting issue involving a specific person in their life (e.g., a friend, roommate, sibling, or partner). We asked participants to engage in an evaluative mode of introspection for the duration of the 6-minute run with the following instructions: “ I would like you to reflect on an aspect of this person's personality or an interaction with them that was mildly upsetting; something that you find difficult to accept and wish were different. As you think about this person and situation, reflect on why this situation is upsetting to you; why the person is the way they are; who caused the situation and why; what has happened in the past to lead up to this point in your relationship; and what might happen going forwards into the future―how things might get worse or better. Think about how you would feel if things with this person were different. As you think about this topic, allow yourself to become fully caught up in your thoughts and emotions. Try to think about what the person and situation means to you, and what aspects are good or bad. Analyze whether you like or dislike the feelings associated with this situation. Analyze why you think the person is the way they are.” Thus, this type of introspection involves reflection on life events within a narrative that links past, present, and future, and the cognitive elaboration and evaluation of those events based on personal preferences. Participants were instructed to keep their eyes closed, but were told that they could open their eyes if they needed a reminder of the instructions (short sentences capturing the essence of the instructions were presented on the screen). Prior to scanning, participants were asked to select the person/situation that they would think about, and were given an overview of the instructions, and asked if they understood how to adopt this mode of introspection. Then just prior to this scanning run, detailed instructions were provided, and participants were asked to engage in this specific type of introspection for the entire duration of the scan. Acceptance-based introspection. We asked participants to engage in a type of introspection inspired by 95  contemporary mindfulness and acceptance-based practices. This type of introspection focuses on cultivating a present-centered awareness, grounded in the acceptance of moment-to-moment viscero-somatic sensations. In this condition, participants were asked to reflect on the same upsetting issue as in the case of the evaluation-based introspection condition, but this time, received the following instructions: “In this condition, please think about the same upsetting issue (same person and situation), but this time, do your best to avoid any type of mental analysis of the situation. Instead, as you hold this person in your mind, try to pay attention to the feelings of your body in the present moment. For example, as you think of this person, notice whether your heart starts to beat faster, or whether your body becomes tense or relaxed, and notice how you are breathing. Try your best to stay with the feelings of your body as they come and go. When thoughts or feelings arise, try to observe them and then let them go. Think of your thoughts and feelings as waves that emerge and then disappear back into the ocean. Try not to analyze your thoughts and feelings as good or bad and try not to think about the cause of the situation or what might happen in the future. Just keep returning to the sensations of your body in the present moment, and allow yourself to accept and experience them fully, whatever they are. Whenever you notice that you have become caught up in a train of thought or you start to analyze your experience, that is okay; just calmly re-focus on how your body is feeling in the present moment. There is no specific goal or purpose other than noticing what you are thinking and feeling from one moment to the next.” Participants were instructed to keep their eyes closed, but were told that they could open their eyes if they needed a reminder of the instructions (short sentences capturing the essence of the instructions were presented on the screen). Prior to scanning, participants were given an overview of the instructions, and asked if they understood how to adopt this mode of introspection. Then just prior to this scanning run, detailed instructions were provided.  Self-reports  Following the scanning session, participants filled in a questionnaire pertaining to their experience during the evaluation and acceptance conditions. Participants rated the following questions on a scale from 1  = not at all, to 7 = a lot / very much: (1) To what extent were you aware of your body and experiencing bodily sensations? (2) To what extent do you feel like you were caught up in your thoughts and feelings? (3) To what extent were you focused on analyzing 96  the causes of the situation? (4) To what extent did you think about what might happen in the future in relation to this situation? (5) To what extent did you judge your thoughts/feelings as good or bad? (6) How difficult was it to follow the instructions and evaluate/accept your thoughts and feelings? Participants also rated their mood during the condition on a 7-point scale from 1 = very unpleasant to 7 = very pleasant.  MRI data acquisition fMRI data were collected using a 3.0-Tesla Philips Intera MRI scanner (Best, Netherlands) with an 8-channel phased array head coil with parallel imaging capability (SENSE).  Head movement was restricted using foam padding around the head.  T2*-weighted functional images were acquired parallel to the anterior commissure/posterior commissure (AC/PC) line using a single shot gradient echo-planar sequence (repetition time, TR = 2 s; TE = 30 ms; flip angle, FA = 90°; field of view, FOV = 240 mm; matrix size = 80 × 80; SENSE factor = 1.0).  Thirty-six interleaved axial slices covering the whole brain were acquired (3-mm thick with 1-mm skip).  Each session was six minutes in length, during which 180 functional volumes were acquired. Data collected during the first 4 TRs were discarded to allow for T1 equilibration effects. Before functional imaging, a high resolution T1-weighted structural image was acquired (170 axial slices; TR = 7.7 ms; TE = 3.6 ms; FOV = 256 mm; matrix size = 256 × 256; voxel size = 1 x 1 x 1 mm; FA = 8°). Total scan time was ~ 60 minutes. Head motion was minimized using a pillow, and scanner noise was minimized with earplugs.  Preprocessing  Image preprocessing and analysis were conducted with Statistical Parametric Mapping (SPM8, University College London, London, UK; http://www.fil.ion.ucl.ac.uk/spm/software/spm8). The time-series data were slice-time corrected (to the middle slice), realigned to the first volume to correct for between-scan motion (using a 6 parameter rigid body transformation), and coregistered with the T1-weighted structural image. The T1 image was bias-corrected and segmented using template (ICBM) tissue probability maps for gray/white matter and CSF.  Parameters obtained from this step were subsequently applied to the functional (re-sampled to 3 mm3 voxels) and structural (re-sampled to 1 mm3 voxels) data during normalization to MNI 97  space. The data were spatially-smoothed using an 6-mm3 full-width at half-maximum Gaussian kernel to reduce the impact of inter-subject variability in brain anatomy.    To address the spurious correlations in resting-state networks caused by head motion, we identified problematic time points during the scan using Artifact Detection Tools (ART, www.nitrc.org/projects/artifact_detect/). Images were specified as outliers according to the following criteria: translational head displacement greater than 0.5 mm from the previous frame, or rotational displacement greater than .02 radians from the previous frame, or global signal intensity > 4 standard deviations above the mean signal for that session. Outlier images were not deleted from the time series, but rather, modeled in the first level general linear model (GLM) in order to keep intact the temporal structure of the data. Each outlier was represented by a single regressor in the GLM, with a 1 for the outlier time point and 0 elsewhere.   Using the 'CONN' software (Whitfield-Gabrieli & Nieto-Castanon, 2012), physiological and other spurious sources of noise were estimated and regressed out using the anatomical CompCor method (Behzadi et al., 2007). Global signal regression was not used due to fact that it mathematically introduces negative correlations (Murphy et al., 2009). The normalized anatomical image for each participant was segmented into white matter (WM), gray matter, and CSF masks using SPM8. To minimize partial voluming with gray matter, the WM and CSF masks were eroded by one voxel. The eroded WM and CSF masks were then used as noise ROIs. Signals from the WM and CSF noise ROIs were extracted from the unsmoothed functional volumes to avoid additional risk of contaminating WM and CSF signals with gray matter signals. The following nuisance variables were regressed out: three principal components of the signals from the WM and CSF noise ROIs; head motion parameters (three rotation and three translation parameters) along with their first-order temporal derivatives; each artifact outlier image; linear trends. A band-pass filter (0.009 Hz < f < 0.10 Hz) was simultaneously applied to the BOLD time series during this step.  Network nodes  We created a set of 255 nodes based on a fusion of the Yeo (Yeo et al., 2011) and Gordon (Gordon et al., 2014) parcellations. From a set of 114 ROIs created by Yeo and colleagues based on their 17-network parcellation (Krienen et al., 2014; Yeo et al., 2015), we created 5mm radius 98  spherical ROIs around the center of mass coordinates. To provide whole-brain coverage we then created additional ROIs based on the Gordon parcellation. Again, we created 5mm radius spherical ROIs around the center of mass coordinates. Finally, we created 4mm spherical ROIs to cover subcortical regions: the striatum (Choi, Yeo, & Buckner, 2012); amygdala (Lindquist et al., 2016); hippocampus and retrosplenial cortex (Spreng et al., 2009; Zeidman, Mullally, & Maguire, 2015); brainstem nuclei (Bar et al., 2016; Garfinkel et al., 2016). For each participant, we extracted the mean timeseries from unsmoothed data for each ROI, and then the residual timeseries (following nuisance regression) was used to compute condition-specific correlation matrices.  Graph analysis: global properties  We computed several graph metrics to look for potential differences in overall network topology during acceptance and evaluation, including mean clustering, global efficiency, and mean degree (Fornito et al., 2016; Rubinov & Sporns, 2010). Mean degree,    , provides a measure of network density and can be used to determine the  wiring cost―how expensive it is to build the network (Latora and Marchiori, 2003)―and is defined as the mean number of edges for each node. It is computed as:                where   , is the number of edges (i.e., correlations > 0) between node i and all other nodes. The weighted clustering coefficient,   , provides a measure of network segregation by quantifying the potential for communication within the immediate neighborhood of a node. Specifically, it is defined as the proportion of neighbors around node i that are also interconnected, and is computed as:                        where     is the number of triangles around node i (i.e., a set of three nodes that are all interconnected), normalized by their intensity (edge weight), and    is the degree (total number of connections) of node i. The mean weighted clustering coefficient for the overall network is 99  simply the average of all   values across nodes in the network. Furthermore, network-level clustering can be ascertained by averaging   values of nodes within each network delineated by the community detection algorithm. Global efficiency, Eglob, provides a measure of network integration (i.e., the capacity for parallel information transfer in the network), and is defined as the average inverse shortest path distance between each pair of nodes Marchiori, 2001, 2003; Achard and Bullmore, 2007. Thus, a network with high global efficiency is one in which it is possible to define a path between any pair of nodes with few intervening nodes.  It is computed as:                                  where        is the shortest weighted path length (distance) between nodes i and j. In a weighted network, path length is computed as the inverse of weight (thus, stronger edge weights translate into shorter path distances). All graph metrics were computed using the Brain Connectivity Toolbox (Rubinov & Sporns, 2010) on the weighted graphs (Fisher z-transformed correlation coefficients) of individual participants, thresholded to retain only positive edge weights. Graph metrics were computed across a range of thresholds (z(r) = .00 to z(r) = .99 in .01-step increments) to demonstrate the lack of dependence upon thresholding. Community detection To delineate the network structure we used the Louvain algorithm (43) implemented with the brain connectivity toolbox (https://sites.google.com/site/bctnet/) (Rubinov & Sporns, 2010) which places nodes into non-overlapping networks. This algorithm finds partitions optimizing the modularity value, Q, by grouping nodes into networks that maximize intra-modular and minimize inter-modular connections (Newman, 2004). The modularity value for weighted graphs,   , is computed as follows:                                 where     is the edge weight between nodes i and j,    is the sum of all weights in the graph,     is the weighted degree of node i, and   is a module containing node i.       is =1 if nodes i 100  and j belong to the same module, and = 0 otherwise. The modularity of a partition is a scalar value between -1 and 1 that quantifies the strength of within-module edges relative to the strength of between-module edges. We submitted graphs thresholded to retain only positive weights to the Louvain algorithm in a two step approach following prior work (Dwyer et al., 2014; Fornito et al., 2012). First, a consensus partition was obtained for each participant by iterating the Louvain algorithm 1000 times on the individual participant correlation matrix. For each iteration, we constructed a 255×255 co-classification matrix with entries of 1 or 0 depending on whether nodes i and j were assigned to the same or different modules. From the set of matrices, we computed the probability that nodes i and j belonged to the same module. The community detection algorithm was then applied to this co-classification matrix to provide a final set of community assignments at the individual level. In the second step, we derived a group-level consensus partition. We constructed a co-classification matrix on the final individual level community assignments and then constructed a group-level co-classification matrix by determining the probability that each pair of nodes were co-classified in the same module across participants. We then ran the Louvain algorithm on the group co-classification matrix 1,000 times and took the partition associated with the highest Q value as the optimal network partition for the group. The number of communities that are detected with this algorithm can be altered with a resolution parameter, gamma (Reichardt & Bornholdt, 2006). We set gamma such that ~ 17 networks were detected in order to identify network structure similar to the parcellation of Yeo et al. (2011). The final partition included 19 networks. Six were composed of 5 or fewer nodes, and were considered undefined. We adjusted the final partition based on fined-grained analyses in prior work: (i) two undefined nodes―the left and right parahippocampal gyrus―were added to DNMTL network based on (Andrews-Hanna, Reidler, Sepulcre, et al., 2010; Yeo et al., 2011); (ii) one network included many DN regions and was split into Core and dorsomedial prefrontal subsystems based on (Andrews-Hanna, Reidler, Sepulcre, et al., 2010); (iii) one network included many FPCN regions and was split into anterior and posterior subsystems based on Dixon et al in prep; (iv) one network included many nodes that were split into a visual network and a posterior attention network based on (Yeo et al., 2011); and (v) the visual network contained a node near the premotor cortex that was removed given that it is not a canonical visual region and because it exhibited relatively low correlations with other nodes in this network. After these adjustments, there were 15 networks in total. Notably, this partition was 101  derived from resting state data, thus providing an unbiased picture of network structure that we used to compare network organization during evaluation and acceptance.  Rich club analysis  Rich club structure refers to the idea that nodes that are highly connected are themselves strongly interlinked, forming a club of richly connected nodes. We calculated the weighted rich club coefficient,      , (Opsahl, Colizza, Panzarasa, & Ramasco, 2008; van den Heuvel & Sporns, 2011) for each participant and condition.       was computed for a range of k values as:                            where    is the sum of weights for a subgraph composed of nodes with a degree larger than k,     is the total number of edges in this subgraph, and        is a vector of node weights from the entire network ranked by strength. Thus, the sum of weights for a subset of nodes with a k exceeding the cut-off value is compared against the sum of weights for the same number of edges, but taking the highest weighted edges in the network. This measure thereby determines whether the strongest connections occur between the nodes with most connections. The rich club coefficient is usually normalized with respect to rich club coefficients computed on random networks. Thus, we computed rich club coefficients on 100 randomized networks, which were created by shuffling edge weights while preserving degree distribution. We then computed a normalized rich club coefficient,          at each level of k, by dividing       by the corresponding mean rich club coefficient across the randomized networks,          :                              We plotted group-averaged rich club curves for a range of k, where rich club structure was present for each participant (i.e.,           > 1) (van den Heuvel & Sporns, 2011). Normalized rich club coefficients larger than 1 indicate that highly connected nodes are more strongly interconnected with each other than expected by chance. We then identified putative rich club regions as those nodes with the greatest number of strong connections. For each participant, the acceptance condition graph was thresholded to retain the strongest 5% of all edges, and then the 102  degree was computed for each node. We then computed the average node degree across participants, and took the top 10% of nodes as the rich club. The 26 identified nodes all had k > 19.   Temporal variability of FC patterns We used a sliding window approach to track the temporal evolution of FC patterns within and between networks of interest. Prior work has shown that functionally-relevant connectivity patterns can be isolated from ~ 60 seconds of data (Gonzalez-Castillo et al., 2015; Leonardi & Van De Ville, 2015; Liegeois et al., 2015; Shirer et al., 2012). To examine time-varying connectivity patterns, the data were filtered (0.0167 Hz < f < 0.10 Hz) based on the window size of 60-seconds in order to limit the possibility of detecting spurious temporal fluctuations in FC (Leonardi & Van De Ville, 2015). For each participant, we computed the mean strength of within-network communication (clustering) and between-network communication within each window, thus providing a time-series of FC values. We then computed the variability (SD) of mean within-network clustering and between-network FC across time.  Statistical testing Because network properties may not be normally distributed, all graph metrics were evaluated for significant differences across conditions using non-parametric permutation testing. For global properties and rich club coefficients, we computed the area between the curves by summing the differences in mean values (e.g., mean clustering) between acceptance and evaluation at each threshold (Bassett et al., 2012; van den Heuvel et al., 2013). The area between the curves for each variable was compared against an empirically derived null distribution. To create the null distribution, condition labels were randomized (while ensuring that there was one of each condition for each participant). The area between the curves was then determined for the pseudo conditions. This process was iterated 10,000 times to create a null distribution. Finally, the two-tailed p-value was defined as the proportion of null values that exceeded the true difference (using the absolute value of random and true difference scores). For the network-level clustering analysis, we computed the mean condition difference on graphs that were thresholded to retain positive values and compared the mean difference for each network against an equivalent null 103  distribution. A similar process was used to assess the statistical significance of between-network FC values.   Brain networks  Here, we provide a summary of the brain regions comprising each network identified in the community assignment analysis.  (1) Default network core subsystem (DNCore): bilateral rostromedial prefrontal cortex, bilateral ventromedial prefrontal cortex, bilateral posterior cingulate cortex, bilateral pregenual anterior cingulate cortex, R ventral pregenual anterior cingulate cortex, bilateral posterior inferior parietal lobule, bilateral superior rostrolateral prefrontal cortex, R superior frontal sulcus, R superior temporal sulcus,  (2) Default network dorsomedial prefrontal cortex subsystem (DNDMPFC): bilateral dorsomedial prefrontal cortex 1, bilateral dorsomedial prefrontal cortex 2, bilateral dorsomedial prefrontal cortex 3, L temporoparietal junction, bilateral inferior frontal gyrus, L lateral temporal cortex 1, L lateral temporal cortex 2, bilateral cerebellum crus 1, bilateral temporopolar cortex 1, R temporopolar cortex 2, L posterior superior temporal sulcus, L pre-supplementary motor area.  (3) Default network medial temporal lobe subsystem (DNMTL): bilateral parahippocampal gyrus, bilateral retrosplenial cortex 1, bilateral retrosplenial cortex 2, bilateral ventral posterior inferior parietal lobule, bilateral dorsal posterior cingulate cortex, bilateral ventral precuneus, R superior frontal gyrus, L superior frontal sulcus, L occipitotemporal cortex.  (4) anterior frontoparietal control network (aFPCN): R Cerebellum Crus 2, L posterior middle frontal gyrus (pMFG), R anterior orbitofrontal cortex, L ventral intraparietal sulcus (vIPS), bilateral middle temporal gyrus (MTG), bilateral anterior inferior parietal lobule (aIPL), L posterior superior frontal gyrus, bilateral middle frontal gyrus (9/46), bilateral rostrolateral prefrontal cortex (RLPFC), bilateral pre-supplementary motor area, L superior frontal sulcus, bilateral anterior inferior frontal gyrus, R anterior inferior frontal sulcus. (5) posterior frontoparietal control network (pFPCN). bilateral posterior middle temporal gyrus (pMTG), bilateral mid intraparietal sulcus (mid IPS), bilateral posterior inferior frontal 104  sulcus/inferior frontal junction (pIFS/IFJ), L anterior inferior frontal sulcus, bilateral intraparietal sulcus (IPS), bilateral posterior intraparietal sulcus (pIPS),. (6) Salience network 1 (SN1) bilateral posterior superior temporal gyrus (pSTG), R inferior frontal gyrus1 (IFG), R inferior frontal gyrus2 (IFG), bilateral superior rostrolateral prefrontal cortex, bilateral anterior temporoparietal junction (aTPJ), bilateral superior frontal sulcus (SFS), bilateral mid-DLPFC, bilateral anterior insula 1, bilateral anterior insula 2, bilateral anterior mid-cingulate cortex (aMCC) 1, bilateral aMCC 2, R ventral anterior mid-cingulate cortex, R pre-SMA1, R pre-SMA 2, L ventral posterior mid-cingulate cortex, bilateral mid insula. (7) Salience network 2 (SN2): bilateral mid insula, bilateral dorsal mid insula, bilateral ventral mid insula, bilateral posterior mid cingulate cortex, R ventral posterior mid cingulate cortex, bilateral supplementary motor area, L posterior superior temporal gyrus, R putamen, bilateral auditory cortex, bilateral supramarginal gyrus.  (8) Dorsal Attention Network (DAN) left: L frontal eye fields, L anterior middle temporal region, L anterior intraparietal sulcus, left ventral intraparietal sulcus, bilateral cerebellum lobule 7, bilateral ventral precentral cortex, bilateral precuneus, bilateral posterior superior frontal sulcus.  (9) Dorsal Attention Network (DAN) right: R frontal eye fields, R ventral intraparietal sulcus 1, R ventral intraparietal sulcus 2, bilateral middle temporal area, R occipitotemporal cortex, R cerebellum lobule 5, bilateral fusiform gyrus, bilateral anterior intraparietal sulcus, L cerebellum crus 2.  (10) Visual network: bilateral V1 central, bilateral V1 peripheral, bilateral extrastriate central, bilateral extrastriate peripheral, bilateral lingual gyrus, bilateral lateral occipital complex, bilateral middle occipital gyrus, bilateral posterior fusiform gyrus.  (11) Somatomotor network (SMN) bilateral dorsal somatomotor cortex, bilateral somatomotor cortex, bilateral ventral somatomotor cortex, bilateral posterior insula, bilateral secondary somatosensory cortex, bilateral somatomotor hand1, bilateral somatomotor hand2, bilateral somatomotor hand3, bilateral somatomotor hand4, bilateral somatomotor hand5, bilateral somatomotor mouth,  bilateral superior parietal lobule, bilateral anterior intraparietal sulcus, L cerebellum lobule 5, R ventral premotor cortex.  105  (12) Auditory network: bilateral superior temporal sulcus, R posterior superior temporal gyrus, R posterior superior temporal sulcus, bilateral rostral superior temporal gyrus, R anterior middle temporal area, R lateral temporal cortex.  (13) Affective network 1 (cortical): bilateral medial orbitofrontal cortex, L anterior orbitofrontal cortex, bilateral posterior orbitofrontal cortex, bilateral subgenual anterior cingulate cortex, bilateral lateral orbitofrontal cortex, L ventral pregenual anterior cingulate cortex, bilateral nucleus accumbens, L parabrachial nucleus, R medial temporopolar cortex.  (14) Affective network 2 (subcortical): bilateral amygdala, L putamen1, bilateral putamen2, bilateral anterior hippocampus, bilateral anterior medial temporal lobe.  (15) Thalamic-Brainstem Network: R parabrachial nucleus (PBN), bilateral periacqueductal gray (PAG), bilateral substantia nigra, R ventral tegmental area (VTA), bilateral globus pallidus, bilateral mediodorsal thalamus, bilateral hypothalamus, L dorsal raphe nucleus, bilateral magnus raphe nucleus.  (16) Undefined nodes: bilateral caudate, R dorsal raphe nucleus, bilateral locus coeruleus, L medial temporopolar cortex, R aIPS, L pre-SMA, bilateral premotor cortex, L VTA, bilateral posterior parahippocampal gyrus, bilateral posterior hippocampus, bilateral ventral PCC. Results Subjective reports During separate scans, participants engaged in evaluation or acceptance while reflecting on an upsetting issue involving a significant person in their life. Participants reported different subjective experiences during the two forms of introspection (Table 3). Acceptance relative to evaluation was associated with higher scores on awareness of bodily sensations (paired t-test: p < .001), and lower scores on getting caught up in thought, analyzing thought, thinking about the future, and judging thoughts and emotions (composite evaluative thinking score; paired t-test: p < .001). Interestingly, participants exhibiting a greater reduction in the composite evaluative thinking score during acceptance showed a greater increase in bodily sensations (r = -.63). This suggests that an open and non-judgmental orientation towards inner experience may allow for greater attention to present moment bodily sensations. Mood was significantly lower during 106  evaluation than rest, consistent with the instructions to reflect on an upsetting issue (p <  .001). Notably, mood was higher during acceptance than evaluation (p = .001) and not statistically different from rest (p = .11), suggesting that acceptance offers a buffering effect, protecting against negative affect, even when holding in mind an upsetting issue and experiencing strong bodily sensations.  Table 3. Subjective reports about introspective experience Variable Rest Evaluation Acceptance Body sensations 5.13 4.28 5.70 Mood 5.26 3.78 4.70 Caught up in thought  5.02 3.85 Analyze situation  4.96 2.83 Thoughts of future  4.83 2.57 Judge thoughts  4.09 2.30 Difficulty  3.07 3.87 Note. Items were rated on a 7-point likert scale from 1 = not at all, to 7 = a lot / very much (the scale for mood was: 1 = very unpleasant to 7 = very pleasant)  Global brain network properties To examine network organization during the introspection conditions, we computed graph theoretic measures on weighted, undirected graphs reflecting correlations between 255 nodes. We first looked for differences across the introspective conditions in key characteristics that describe the organization of the global brain network (Rubinov & Sporns, 2010). Clustering reflects the extent to which the immediate neighbors of a given node are also interconnected, and provided a measure of segregated information processing. Global efficiency reflects the extent to which nodes are interconnected through short pathways, and provided a measure of the capacity 107  for information exchange/integration. Degree reflects the average number of connected neighbors a node has, and provided a measure of network density. We found no difference between evaluation and acceptance on any global metric [mean weighted clustering: p = .98; global efficiency: p = .21; mean degree: p = .52) (Figure 17).    Figure 17. Global network metrics computed across a range of correlation thresholds.   Within-network communication  We next identified the modular structure of the brain (Figure 18) using a community detection algorithm (Blondel et al., 2008) that optimizes a modularity value, Q, by grouping nodes into modules that maximize intra-modular connections and minimize inter-modular connections (Newman, 2004). This approach partitioned the brain into 15 modules or networks, similar to prior brain parcellations (Power et al., 2011; Yeo et al., 2011). We then looked for differences across introspective conditions in within-network specialized processing capacity, indexed with the clustering coefficient computed separately for each network. We focused on several networks of interest that ostensibly contribute to processes critical for introspection: anterior frontoparietal control network (aFPCN), which plays a role in metacognitive awareness and the deliberate regulation of attention (Dixon et al., 2014b; Dosenbach et al., 2007; Duncan, 2013; Fleming et al., 2010; McCaig et al., 2011; Miller & Cohen, 2001); default network core subsystem (DNCore), which plays a role in self-referential thinking (Andrews-Hanna, Smallwood, et al., 2014); default network medial temporal lobe subsystem (DNMTL), which plays a role in episodic memory and promoting thought variability (Christoff et al., 2016); thalamic-brainstem network and salience 108  networks 1 and 2 (SN1 and SN2), which contribute to interoceptive awareness (Craig, 2002; Critchley & Harrison, 2013; Damasio & Carvalho, 2013; Farb et al., 2012; Seeley et al., 2007); and affective networks 1 and 2, which contribute to emotional appraisals (Dixon, Thiruchselvam, et al., 2017). We found no differences between conditions in mean clustering for any network [DNCore: p = .30, DNMTL: p = .24, SN1: p = .90, SN2: p = .73, aFPCN: p = .95, affective network 1: p = .51, affective network 2: p = .25, thalamic-brainstem network: p = .82] (Figure 19). This suggests that the strength of within-network specialized processing capacity was similar across introspective conditions.    Figure 18. Nodes and network topology during rest. (A) Nodes grouped into networks based on a community detection algorithm. Abbreviations: DN, default network; DMPFC, dorsomedial prefrontal subsystem; MTL, medial temporal lobe subsystem; aFPCN, anterior frontoparietal control network; pFPCN, posterior frontoparietal control network; SN, salience network; DAN, 109  dorsal attention network; SMN, somatomotor network. (B) Visualization of the network topology. The group-averaged correlation matrix during the resting state was thresholded to retain connections with z(r) > .15, and then submitted to the Kamada–Kawai energy algorithm (Kamada & Kawai, 1989), implemented in Pajek software (De Nooy et al., 2011). This algorithm produces spring-embedded layouts that minimize the geometric distances of nodes based on their topological distances in the graph. Well-connected nodes are pulled towards each other, whereas weakly-connected nodes are pushed apart in a manner that minimizes the total energy of the system. Nodes are colour-coded based on community assignment.   Figure 19. Mean clustering for each of the 15 networks. Error bars represent between subject SEM. Abbreviations: DN, default network; DMPFC, dorsomedial prefrontal subsystem; MTL, medial temporal lobe subsystem; aFPCN, anterior frontoparietal control network; pFPCN, posterior frontoparietal control network; SN, salience network; DAN, dorsal attention network; SMN, somatomotor network; L, left hemisphere; R, right hemisphere.  Flexible reconfiguration of between-network interactions We next examined potential differences in between-network interactions across conditions. We first examined network interactions that may correspond with heightened awareness of moment-to-moment bodily sensations during acceptance. Specifically, we hypothesized that acceptance 110  relative to evaluation would be associated with increased coupling between the aFPCN and the SN1 and SN2, which include the anterior and mid insula, respectively―key regions involved in interoceptive awareness (Craig, 2002; Critchley et al., 2004; Farb et al., 2012). Consistent with this idea, aFPCN-SN2 coupling was stronger during acceptance than evaluation (p = .021) (Figure 20A). Interactions between the aFPCN and SN1 did not vary across conditions (p = .93).   We further hypothesized that the introspective conditions would differ in the nature of interactions between DN subsystems. A recent model (Christoff et al., 2016) suggests that the DNMTL promotes variability in thought by reactivating and combining neural ensembles encoding past experiences. Given that the aim of acceptance is to adopt an open and unconstrained mindset, we hypothesized that this condition would be associated with greater thought variability, and elicit stronger cooperation between the DNCore-DNMTL, reflecting enhanced DNMTL contribution to self-referential thought. Consistent with this idea, we found a trend towards greater DNCore-DNMTL coupling during acceptance compared to evaluation (p = .066). Furthermore, a whole-brain analysis revealed extensive FC between a seed region in the MTL and regions of the DNCore during acceptance, but minimal FC with the DNCore during evaluation (Figure 20B). We further hypothesized that if the DNMTL plays a role in promoting unconstrained thought, then individual differences in the strength of FC within the DNMTL may correlate with the strength of DNMTL-DNCore coupling. We found a positive relationship in both conditions, however, the correlation was considerably stronger during acceptance (r = .65) than evaluation (r = .30) (Figure 20C). These findings suggest that acceptance is associated with enhanced coordination between these DN subsystems.  Finally, we examined coupling strength between the DNCore and salience networks. We expected to observe negative coupling during acceptance, reflecting the aim of experiencing bodily sensations without thinking about them or relating them to the self. The prediction was less clear for the evaluation condition. If evaluative/elaborative thinking in response to an emotional challenge occurs in a manner that is an extension of ongoing bodily sensations then there may be positive coupling between the DNCore and salience networks. On the other hand, if evaluative/elaborative thinking is not grounded in embodied experience, then there may be negative coupling in this condition as well. Consistent with the latter idea, we found negative coupling between the DNCore and salience networks in both conditions, and the strength of 111  negative coupling was not statistically different [DNCore-SN1: p = .78; DNCore-SN2: p = .35]. To summarize, acceptance is associated with reconfiguration of specific large-scale network interactions in line with greater awareness of bodily sensations and an open and unconstrained mode of thinking.    Figure 20. Reconfiguration of between-network coupling patterns. (A) Mean between-network FC. (B) Whole-brain seed maps (Z > 3.1, p < .05 FWE cluster corrected) showing voxels within the DNCore (black borders) that are correlated with a seed region in the medial temporal lobe (parahippocampal gyrus). Colour bar represents t-scores. (C) Scatter plots demonstrating that 112  individual differences in the strength of DNMTL within-network FC are correlated with the strength of DNCore-DNMTL coupling, and this relationship is stronger during acceptance.   Rich club  We next examined the hub structure of the brain. Hub regions support integrative processing by serving as bridges that interconnect nodes from spatially segregated networks (Power, Schlaggar, Lessov-Schlaggar, & Petersen, 2013; van den Heuvel & Hulshoff Pol, 2010). One index of the brain's overall hub organization is the “rich club” structure―a tendency for highly connected hub nodes to be preferentially connected to each other rather than with nodes with fewer connections. Rich club structure provides a stable core associated with integrative communication across different functional modules and may preferentially contribute to interoception rather than exteroception (Gollo et al., 2015; van den Heuvel & Sporns, 2011). Accordingly, we predicted an increase in rich club structure during acceptance, reflecting greater global integrative processing while individuals monitor thoughts and viscerosomatic sensations in an open and unconstrained manner. The results demonstrated rich club structure in both introspection conditions, evidenced by normalized rich club coefficients greater than 1 across a range of k, from 107-153. Notably, rich club coefficients were larger for acceptance than evaluation (p = .05; range k = 107 to k = 141) (Figure 21A). To delineate the brain regions that participate in the rich club, we identified nodes that were strongly interconnected with a large number of other nodes (i.e., nodes with a high degree; k > 19 at 5% network edge density). This revealed a collection of nodes spanning the DN, FPCN, and salience network (Figure 21B). These regions align with prior work on the brain's structural core (Hagmann et al., 2008), rich club structure (van den Heuvel & Sporns, 2011), connector hubs (Power et al., 2013), and with the “multiple demand” system that is frequently recruited across diverse tasks (Duncan, 2010).   113   Figure 21. Rich club structure during evaluation and acceptance. (A) The figure shows rich-club structure (normalized rich-club coefficient > 1) for a range of k from 107 to 153. Rich club values were greater during acceptance than evaluation, suggesting increased tendency for hub regions to be preferentially interconnected. (B) Rich club nodes, colour-coded based on module assignment: L posterior cingulate cortex (PCC), L posterior inferior parietal lobule (pIPL), R dorsomedial prefrontal cortex (DMPFC), L lateral temporal cortex (LTC), R medial temporal lobe (MTL), bilateral posterior middle frontal gyrus (pMFG), L middle temporal gyrus (MTG), L posterior superior frontal gyrus (pSFG), bilateral rostrolateral prefrontal cortex (RLPFC), bilateral pre-supplementary motor area (SMA), L superior frontal sulcus (SFS), bilateral anterior inferior frontal gyrus (aIFG), L ventral intraparietal sulcus (vIPS), L posterior middle temporal gyrus (pMTG), L posterior inferior frontal sulcus/inferior frontal junction (pIFS/IFJ), posterior intraparietal sulcus (pIPS), L ventral precentral cortex (PrCv), L Cerebellum Crus 2. 114   Temporal variability of FC patterns  We hypothesized that the mindset engendered by acceptance would lead to greater variability in mental content across time. We thus predicted that this condition would be associated with a corresponding increase in the temporal variability of FC patterns. In particular, we examined the variability of clustering within the DNMTL, and the variability of DNCore-DNMTL coupling strength. Using a sliding-window approach, we tracked FC patterns as they evolved across time, and operationalized variability as the standard deviation (SD) of FC strength across time. Consistent with our prediction, there was greater temporal variability in DNCore-DNMTL coupling strength during acceptance than evaluation (p = .004; SD evaluation: .068; SD acceptance: .085). There was a no difference across conditions in temporal variability of DNMTL clustering (p = .56; SD evaluation: .014; SD acceptance: .013). These results suggest a correspondence between the presumed cognitive variability that occurs during acceptance and neural variability as operationalized here.    High versus low acceptors  Subjective reports revealed that participants differed in their ability to adopt the acceptance mindset. To examine the relationship between subjective experience and network configuration, we divided participants into two groups, “high acceptors” and “low acceptors” based on a median split of their reverse-coded evaluative thinking score. High relative to low acceptors exhibited changes in global network properties including reduced clustering (p = .032) and reduced global efficiency (p = .027), and a trend towards reduced mean degree (p = .083) (Figure 23A). Reduced clustering was observed in many of our networks of interest at a significant or trend level [DNCore : p = .011; DNMTL: p = .074; aFPCN: p = .086; SN1: p = .017; SN2: p = .17; affective 1: p = .032; affective 2: p =.003; thalamic-brainstem: p =.016] (Figure 22B). In contrast, there was no difference between high and low acceptors in rich club structure (p = .86) (Figure 23C) or task-relevant between-network interactions (p's > .55) (Figure 22D). These findings suggest that high acceptors exhibited decreased brain modularity across cortical and subcortical systems and a decreased number of short paths between nodes throughout the brain, but preserved communication between hub regions and between task-relevant networks.  115    Figure 22. Network properties in high acceptors and low acceptors during the acceptance condition. (A) High relative to low acceptors demonstrated reduced clustering and global efficiency, and a trend towards reduced degree. (B) Mean weighted clustering for each network. (C) There was no difference between high and low acceptors in rich club structure. (D) There was no difference in relevant between-network interactions. Error bars represent between subject SEM.  116  Discussion Our results provide novel evidence that complex subjective experiences during internally-oriented attention can be linked to the dynamic reorganization of network properties. Subjective reports demonstrated that acceptance-based introspection relative to evaluation-based introspection was associated with greater awareness of bodily sensations, reduced judgmental thought, reduced past and future thinking, and a buffering effect on mood. These differences in subjective experience were accompanied by changes in specific between-network interactions, namely, increased coupling between the aFPCN and salience 2 network, and increased coupling between the DNMTL and DNcore. Acceptance was also associated with greater temporal variability in the strength of DNMTL-DNcore coupling. We also found increased rich club structure, an index of integrative processing capacity. Finally, high relative to low acceptors exhibited reduced clustering and global efficiency, but no change in hub structure or task-relevant between-network interactions, suggesting that acceptance may be linked to a less modular network organization. Our findings complement a growing literature documenting systematic changes in network patterns across time and context, and linking such changes to behavioral task performance (Bassett et al., 2011; Braun et al., 2015; Dwyer et al., 2014; Fornito et al., 2012; Shine, Bissett, et al., 2016).  Enhanced coordination between frontoparietal and interoceptive networks In contrast to evaluation which involves the cognitive elaboration of arising thoughts and emotions, the goal of acceptance is to keep attention anchored on moment-to-moment fluctuations in bodily sensations, without ascribing meaning or value-judgments to those sensations. Accordingly, we expected that network reconfiguration during acceptance would involve brain regions that support interoceptive awareness. Interoception relies on ascending pathways that have been well-mapped out in non-human primates using invasive tracing studies. Signals related to the viscera, tissue damage, temperature, and metabolic processes are conveyed via the lamina I spinothalamic tract and vagus nerve to brainstem centers including the nucleus of the solitary tract (Craig, 2002; Critchley & Harrison, 2013; Damasio & Carvalho, 2013). From there, information is passed to the parabrachial nucleus and periacqueductal gray (PAG), which provide an initial integrative representation of the internal state of the body (Damasio & Carvalho, 2013). These regions project to subcortical structures important for regulating 117  homeostasis including the ventromedial thalamus, hypothalamus, and amygdala. Finally, these regions send efferent projections conveying interoceptive signals to a limited number of cortical centers including the insula (Craig, 2002; Critchley et al., 2004; Farb et al., 2012). We thus hypothesized that acceptance would be associated with enhanced coupling between the aFPCN, which plays a role in metacognitive awareness and the deliberate control of attention, and the salience networks, which include parts of the insula (Kleckner, 2017; Seeley et al., 2007). Consistent with this, acceptance was associated with increased coupling between the aFPCN and salience network 2 (SN2) which includes the mid insular cortex. The mid and posterior sectors of the insula have been identified as primary interoceptive cortex given activation patterns that correlate with fluctuations in objectively measured bodily signals (e.g., respiratory, cardiac, pain) (Craig, 2002; Farb et al., 2012). Other regions in the SN2 network include the mid-cingulate cortex and supramarginal gyrus (i.e., the most anterior part of the temporoparietal junction), which contribute to action tendencies (Vogt, 2005) and the conscious experience of the sense of body location in space (Blanke, 2012). The aFPCN and SN2 were negatively correlated during evaluation, suggesting that evaluative reactivity to an emotional challenge may paradoxically result in less awareness of bodily signals. On the other hand, these networks were uncorrelated during acceptance which may suggest that these networks are going in and out of phase. Indeed, on a finer temporal scale, the aFPCN and SN2 exhibited positive coupling in a greater percentage of time-windows during acceptance (22%) than during evaluation (12%) suggesting an increase in the amount of time during which these networks are coordinated, and potentially enabling metacognitive awareness of bodily sensations.   Prior work has shown that FPCN and salience network regions normally deactivate during sadness provocation, but not after meditation training (Farb et al., 2010). This suggests that individuals may often attend to their thoughts and judgments about bodily feelings rather than attend to bodily feelings themselves, unless specifically directed to do so. Moreover, the fact the evaluation was associated with negative coupling between the DNCore and salience networks provides new evidence that evaluative/elaborative thinking during an emotional challenge is not an extension of interoceptive processing, but rather, is divorced from concrete bodily sensations. On the other hand, acceptance was associated with greater reported bodily sensations and increased recruitment of interoceptive networks, suggesting that its protective effect on mood during emotional challenges does not reflect a denial or blunting of affect, but 118  rather, a shift in orientation towards experience based on a willingness to tolerate uncomfortable sensations coupled with a letting go of the tendency to figure out why one is experiencing negative feelings, and relating those feelings to one's sense of self (Farb et al., 2010).  Enhanced coordination between DN subsystems related to thought variability Acceptance is characterized by an open and unconstrained mindset in which thoughts and feelings are allowed to spontaneously rise and fall. In this way, acceptance may facilitate variability in mental content, given the one is not attempting to fit thoughts and feelings into a coherent narrative. The DNMTL contributes to episodic memory (Addis et al., 2007; Andrews-Hanna, Saxe, et al., 2014) and scene construction (Hassabis, Kumaran, & Maguire, 2007), and demonstrates increased activation just prior to the arising of spontaneous thoughts (Ellamil et al., 2016). The DNMTL may promote thought variability through its role in retrieving and recombining associative knowledge (Christoff et al., 2016). Accordingly, we hypothesized that periods of increased coupling between the DNMTL and DNCore  may correspond to less evaluative and more free flowing and spontaneous thought dynamics. In line with this, we found that acceptance relative to evaluation was associated stronger DNMTL-DNCore coupling. Additionally, individual differences in the strength of DNMTL within-network FC demonstrated a stronger correlation with the strength of DNMTL-DNCore coupling during acceptance. While our findings do not speak to directionality of information flow, one possibility is that greater inputs from the DNMTL shift DNCore processing away from self-evaluative narrative-based thinking to an internal monitoring of spontaneously arising and dissolving thoughts. On the other hand reduced communication between the DN subsystems during evaluation may allow the DNCore to promote elaborative self-related thinking and to some extent suppress spontaneously activated mnemonic content in the DNMTL which may otherwise trigger new streams of thought. In extreme cases, de-coupling between the DNMTL and DNCore may contribute to the inflexible thinking patterns that characterize mood disorders (Beck, 1991). While prior work has found reduced activation of the default network in meditators relative to novices during acceptance (Farb et al., 2007), here, we show a divergent pattern at the level of network communication, with enhanced FC across nodes in different subsystems of the DN. This finding reinforces the notion that activation magnitude may often be orthogonal to network communication patterns (Dixon, Andrews-Hanna, et al., 2017; Murphy et al., 2016).  119   We also found that acceptance was associated with a significant increase in the variability of DNMTL-DNCore coupling strength across time. A growing number of studies have investigated temporal variability in local brain activation magnitude and inter-regional coupling patterns and suggest that it may be a signature of cognitive flexibility (Allen et al., 2014; Bassett et al., 2011; Braun et al., 2015; Dixon, Andrews-Hanna, et al., 2017; Hutchison, Womelsdorf, Gati, et al., 2013; Mitra et al., 2015; Zalesky et al., 2014). Temporal variability may reflect moment-to-moment changes in mental content, or a process of sampling diverse neurocognitive states, akin to a tennis player swaying back and forth to maximize readiness before an opponent's serve (Deco et al., 2011). Thus, one possibility is that increased variability of DNMTL-DNCore coupling strength corresponds with the flexible and open mindset encouraged by an acceptance-mode of introspection, which presumably facilitates variability in mental content. Moreover, one intriguing hypothesis is that increased temporal variability of DNMTL-DNCore coupling strength may be a signature of adaptive processing during an emotional challenge that protects against ruminative thinking.  Enhanced rich club structure Prior work has demonstrated the existence of topologically central “hub” regions that share connections with many other nodes, thus providing an important infrastructure for the integration of information across spatially distributed specialized processing modules (Power et al., 2013; van den Heuvel & Sporns, 2011, 2013). Notably, these regions also tend to be preferentially interconnected among themselves―the so called “rich club” structure (van den Heuvel & Sporns, 2011). We found that acceptance relative to evaluation was associated with increased rich club structure. Consistent with prior work (Hagmann et al., 2008; Power et al., 2013; van den Heuvel & Sporns, 2011), we identified rich club nodes in association networks, most notably, frontoparietal, default, and salience networks. The rich club nodes exhibited a specific organization, with several nodes that belong to different networks being in close spatial proximity within anterior prefrontal and inferior parietal cortices. This spatial organization may facilitate information integration across diverse functional systems. One possibility is that enhanced rich club structure during acceptance enables attention to flexibly monitor  the diverse sensory, interoceptive, and mnemonic signals that arise from moment-to-moment. Moreover, the adaptive coding properties of frontoparietal neurons (Duncan, 2001; Stokes et al., 2013) may 120  contribute to the effective engagement and disengagement of attention to thoughts and emotions. Activity patterns in the lateral prefrontal cortex during different trial events (e.g., cue, delay, target) are nearly orthogonal with many cells coding specific information at each stage, but showing little correspondence in what is coded across stages (Stokes et al., 2013). Thus, the information represented across the population of frontoparietal neurons is constantly reorganized to reflect currently relevant information (Duncan, 2013). This rapid and flexible coding may contribute to the ability to notice and then let go of arising mental content. Finally, recent evidence suggests that the activity dynamics of the rich club may be ideally suited for a role in interoceptive awareness (Gollo et al., 2015), which is emphasized during acceptance.   A recent study found that LSD administration was associated with reduced rich club structure (Tagliazucchi et al., 2016). While psychedelic experience shares similarities with meditation and acceptance-based practices including a heightened sense of present-moment awareness, there are also important differences. LSD is often associated with a sense of ego-dissolution and changes in the sense of boundary between self and environment, whereas this is not a typical experience in acceptance/meditation, except perhaps in expert practitioners. One possibility is that the reduced rich club structure in LSD is specifically related to the breakdown of self-identity and bodily boundaries (Tagliazucchi et al., 2016). Future work directly contrasting network organization during acceptance/meditation and psychedelic states may yield new insights into the network properties underlying specific changes in subjective experience.   High versus low acceptors Considerable evidence suggests that the brain exhibits small-world organization (Bassett & Bullmore, 2006; Bassett & Bullmore, 2016; Bullmore & Sporns, 2009) characterized by high local neighborhood clustering together with sufficient long-distance connections to support global communication (Latora & Marchiori, 2001; Watts & Strogatz, 1998). Our findings suggest that acceptance may be a unique state that departs, to some extent, from typical small-world organization. Participants with high acceptance scores exhibited decreased clustering and a decreased number of short paths across the brain, but preserved preferential communication between hub regions, and no difference in coupling strength between task-relevant networks. One interpretation is that the network topology in high acceptors reflects a specific reorganization of information flow to and among the brain's most connected regions, which may 121  support the open monitoring of mental activity. Moreover, diminished global efficiency yet preserved hub structure suggests that the network topology of high acceptors may reflect a highly efficient state in which task-irrelevant connections are diminished. In other words, there may be a selective dampening of short paths that are irrelevant for instantiating this type of mental activity. However, at present this remains speculative and requires further investigation. Nevertheless, the present findings indicate a novel link between the effective implementation of acceptance and diminished network clustering and number of short paths across the brain.  Limitations A few limitation should be noted. First, we collected reports after scanning rather than online during the introspection conditions, which may potentially lead to misremembering and erroneous reports. However, the agreement between subjective changes and observed differences in specific brain network interactions argues against this idea as a major factor. Another potential issue is that subjective reports can be prone to expectancy or demand effects, and there is no way to objectively verify that individuals were able to adopt the distinct modes of introspection in accordance with the instructions. This is a limitation that cannot be avoided because these inner experiences do not have physiological or behavioral correlates that can be objectively measured. For example, while acceptance may sometimes lead to a shorter or less intense experience of emotion due to the lack of cognitive reactivity, this need not always be the case. Allowing emotion to be fully experienced in the body without judgment may in some cases heighten physiological or behavioral reactions. Thus, physiological/behavioral responses can be independent of the subjective experience of those responses. A final limitation is that the acceptance condition always followed the evaluation condition. We held the order constant because evaluation appears to be the default introspective style of most individuals (Farb et al., 2007) and we hypothesized that having acceptance come before evaluation may lead to carryover effects due to the challenge of implementing this type of introspection. Notably, two key findings argue against the idea that observed network differences were due to the order of the conditions. First, there were no differences in global network properties, which may be expected if a general condition order effect such as fatigue was driving the effects. Second, we found specific increases and decreases in the strength of theoretically-relevant network interactions. Such specificity renders it unlikely that non-specific condition order effects can explain or findings. 122  Conclusions Our findings reveal dynamic changes in network topology across evaluation-based and acceptance-based introspection. This suggests that network organization is tightly linked to subjective experience during internally-oriented attention. In particular, acceptance is associated with reconfiguration of specific large-scale network interactions in line with greater awareness of bodily sensations and an open and unconstrained mode of thinking, as well as increased rich club structure, which may facilitate a reflective monitoring of diverse exteroceptive and interoceptive signals as they change from moment to moment. More broadly, our findings support the notion that network organization exhibits structured changes across time and context, and may have implications for understanding neuropsychiatric disorders that involve disturbances in affect and introspective processes.       123  CHAPTER 5 - GENERAL SUMMARY AND DISCUSSION  Summary of main findings Attention is a fundamental property of cognition enabling some aspects of the inner or outer environment to be highlighted at the expense of others. This creates a manageable information flow and facilitates memory, decision making, planning, and self-reflection, by boosting the processing of only the most relevant signals at any given moment. While much is known about the neural basis of attention, much of this work relied on task activation patterns or static measures of functional connectivity. The present research used graph theoretic tools to delineate temporal and contextual changes in large-scale brain properties to provide novel insights into the nature of attention at cognitive and neural levels of analysis. In chapter 2, we found that the DN and DAN are not intrinsically anticorrelated. Rather, their relationship varied across time and context, arguing against the notion of an inherent competition between perceptual and conceptual information processing. In chapter 3, we examined the functional organization of the FPCN, which is known to play a key role in the deliberate regulation of attention. While the FPCN often appears to function as a domain general control network, we found evidence in favour of a finer-level of organization, with distinct anterior and posterior subsystems related to the DN and DAN, respectively. This FPCN fractionation was present across different contexts, but varied in strength, and was critical for understanding how the FPCN relates to changes across time in the strength of DN and DAN clustering. This provides new information about the network properties that may underlie dynamic changes in the deliberate control of perceptual and conceptual attention. In chapter 4, we examined different types of internally oriented attention, and found that individuals can adopt distinct modes of introspection with unique underlying network configurations. In particular, acceptance-based introspection which involves an open and unconstrained mindset and enhanced interoceptive awareness was associated with increased coupling strength between the aFPCN and SN2 and between the DNMTL and DNCore, as well as enhanced rich club structure. Additionally, acceptance was associated with greater temporal variability in DNMTL-DNCore coupling strength, consistent with the presumed increase in the variability of mental content. This suggests that distinct introspective contexts can be linked to network reorganization. Together, this work provides compelling evidence in favour of a dynamic view of network organization underlying attention, with flexible shifts in network 124  interactions across time and context. In this chapter, I expand upon the implications of these findings. Network organization is flexible In the last decade, there has been considerable focus on identifying an intrinsic functional network architecture of the brain (M. Fox & Raichle, 2007). It is believed that the networks and interactions observed during rest should be largely invariant to task demands and cognitive state (Honey et al., 2009; Margulies et al., 2009; Raichle, 2009; Smith et al., 2009; van den Heuvel & Hulshoff Pol, 2010; Van Dijk et al., 2010). In agreement with this view, there does appear to be a core network backbone that persists across different tasks (Cole, Bassett, et al., 2014; Krienen et al., 2014). This perspective has had a major impact on the field and has fueled countless studies looking for deviations from this intrinsic architecture that may characterize brain development, clinical conditions, and cognitive decline during aging. One of the putative markers of a healthy intrinsic architecture is DN-DAN anticorrelation, which ostensibly reflects an inherent competition between external perceptual and internal conceptual processing (Fox et al., 2005; Kelly et al., 2008). However, accumulating evidence suggests that network organization is more dynamic than previously thought, with documented changes across context and time (Allen et al., 2014; Buckner et al., 2013; Fornito et al., 2012; Gonzalez-Castillo et al., 2015; Hutchison, Womelsdorf, Gati, et al., 2013; Krienen et al., 2014; Milazzo et al., 2014; Mitra et al., 2015; Shine, Bissett, et al., 2016; Shirer et al., 2012; Spreng et al., 2010; Zalesky et al., 2014). Our findings build upon this work in relation to the network basis of attention.   In chapter 2 we found that DN-DAN interactions varied across different task contexts and also varied substantially across time. Indeed, there were periods of anticorrelation that alternated with periods of positive correlation. These context- and time-varying interactions suggest that there is no inherent competition between the DN and DAN, and by extension, perceptual and conceptual processing. Rather, the nature of their interaction may depend on what is currently salient. Thus, in contrast to the mutually suppressive effect that occurs between two objects within the receptive field of a single neuron (Desimone & Duncan, 1995), there does not appear to be an analogous inherent competition at a broader level between perceptual and conceptual processing. Demonstrating the variability of DN-DAN coupling has important 125  theoretical implications as it suggests that DN-DAN anticorrelation is not clear signature of “healthy” functioning.   Our findings also reveal that a dynamic perspective is critical for understanding how the FPCN―a critical network involved in the deliberate attentional control―relates to the DN and DAN. In chapters 2 and 3 we found changes across time in the DN and DAN that occur in concert with temporal fluctuations in coupling with the FPCN. Specifically, in chapter 2, we demonstrated that periods of stronger negative coupling between the FPCN and DAN were associated with stronger negative coupling between the DAN and DN. In chapter 3, we fractionated the FPCN and demonstrated a relationship between temporal fluctuations in aFPCN-DN coupling and DN clustering, and a relationship between temporal fluctuations in pFPCN-DAN coupling and DAN clustering. Thus, our findings demonstrate structured changes in network interactions across time that are not captured with static measures of FC. One possibility is that these temporal shifts in the strength of FPCN interactions reflect dynamic changes in the deployment of deliberate control and the strength of competitive weights assigned to perceptual versus conceptual information. However, a limitation of the present research is the lack of behavioral data and the possibility of linking these dynamic network properties to objective indices of attention.     Finally, in chapter 4 we provided evidence of network reorganization during different types of internally-oriented attention. We examined two modes of introspection that differ in the salience of evaluative/elaborative thought versus interoceptive signals. Not only did specific between-network interactions change in strength, but we also found a global change in the hub structure of the brain. Furthermore, dividing our participants into high relative to low acceptors based on subjective reports further revealed that acceptance was associated with diminished clustering, an index of network modularity and segregated information processing. Thus, fundamental network properties reorganize as a function of the focus of attention.   Our findings suggest that a static measurement of FC during the resting state―as done in the majority of studies to date―offers a window into network architecture in that context, but does not reveal brain organization in general. Rather, network organization is flexible and constantly adapting to changes across time and context in the salience of objects, thoughts, and bodily sensations. Accordingly, group comparisons (e.g., healthy adults versus those with 126  depression) based on resting state data alone cannot distinguish between true differences in brain organization from a momentary difference resulting from a divergence in mental state. A more fruitful approach may be to compare groups in flexibility across time or context. Some clinical conditions (e.g., depression) may be associated with diminished flexibility whereas others (e.g., ADHD) may be associated with excessive flexibility.   Additional insights can be gleaned in future work that examines changes in network properties in relation to specific behavioral measures of task performance in different contexts. It will also be important to explore individual differences. Most studies to date have focused on average group-level brain network properties. However, we found considerable individual variability in how network organization changed across contexts in chapter 2. For example, some individuals demonstrated greater negative correlation between the DN and DAN during movie viewing relative to rest, while other individuals demonstrated the opposite pattern. This could have reflected differences in preference or history with the movie, which may have altered the way it was processed. What is salient for one individual may not be for others, so it is critical to account for such variability in network analyses. Unfortunately, our sample was not large enough to provide an in-depth analysis of individual differences.     Attention relies on distributed network configuration  Research on the network basis of attention has remained relatively fragmented. One line of inquiry has focused on the DAN and its role in visual attention (Bisley & Goldberg, 2003; Buschman & Kastner, 2015; Buschman & Miller, 2007; Corbetta & Shulman, 2002; Gottlieb et al., 1998; Moore & Armstrong, 2003). A separate line of work has focused on the SN and suggested that it is the critical network for signaling the salience of internal and external events, based on neuroimaging work revealing activation in a number of contexts during which salient stimuli are presented (Menon & Uddin, 2010; Seeley et al., 2007). A large number of studies on the DN suggest a role in internally-oriented conceptual attention (Andrews-Hanna, Smallwood, et al., 2014; Buckner et al., 2008; Christoff et al., 2016). Finally, the FPCN has been attributed a broad role in the deliberate regulation of attention.   By taking a step back and considering these networks in relation to each other and from a dynamic perspective, it becomes clear that salience emerges from the combined influence of all 127  of these networks. In other words, there is no one network that is serves as a global saliency map. Abundant evidence indicates that the DAN is not a general attention network, but rather, specifically involved in perceptual attention. Thus, the DAN contributes to competitive weights assigned to perceptual signals. Similarly, the poorly named “salience network” is not uniformly activated by salient stimuli; anatomical connections and brain activation patterns suggest a preference for interoceptive signals. Thus, the SN contributes to competitive weights assigned to bodily sensations. The DN is not adequately captured by a functional description in terms of “internal attention”. It can be fractionated into separate subsystems that are preferentially involved in self-referential thought (DNCore), retrieving and recombining learned associations (DNMTL), and social conceptual knowledge (DNDMPFC). Accordingly, these subsystems contribute to competitive weights assigned to these types of content. Finally, we have shown that the FPCN can be fractionated into anterior and posterior subsystems that ostensibly contribute to the deliberate control of internally-oriented conceptual thought and perceptual processing, respectively. The FPCN may thus exert a higher-order weight that modulates the relative importance of processes occurring in other networks based on current task demands. We suggest that attention can be understood in terms of the interactions across networks as they dynamically unfold across time and context. In other words, what is most salient (and attended) at any given moment may depend upon the specific configuration of interactions across all of these networks (and others not covered here). Our findings provide some evidence in favour of this idea. For example, we found simultaneous changes in coupling strength across multiple networks during acceptance relative to evaluation, which likely reflected the shift in the salience of interoceptive signals and the salience of spontaneous versus elaborative thought.   Most studies to date have not considered attention from such an encompassing perspective, given that most paradigms have been designed to examine salience within a single domain (e.g., the relative salience of different visual objects). Yet, a model of dynamic interactions across multiple networks may be the only way to capture the workings of attention in more naturalistic contexts in which multiple factors simultaneously weigh in on the relevance of a multitude of interoceptive and interoceptive signals.      128  Conclusions The research presented herein offers a window into the contextual and temporal variability in network interactions that are related to attention. Our findings highlight the necessity of considering attention within a dynamic network framework. The next step will be to link the dynamic network properties observed here with experience sampling and behavioral assays of attention in order to provide a more direct association between the neural and cognitive levels of attention. In sum, a dynamic network neuroscience approach using graph theory will play an essential role in delineating brain organization underlying contextual and temporal changes in attention.          129  REFERENCES  Addis, D. R., Wong, A. T., & Schacter, D. L. (2007). Remembering the past and imagining the future: common and distinct neural substrates during event construction and elaboration. Neuropsychologia, 45(7), 1363-1377. Allen, E., Damaraju, E., Plis, S. M., Erhardt, E. B., Eichele, T., & Calhoun, V. D. (2014). Tracking whole-brain connectivity dynamics in the resting state. Cereb Cortex, 24(3), 663-676. Allen, M., Smallwood, J., Christensen, J., Gramm, D., Rasmussen, B., Jensen, C. G., et al. (2013). The balanced mind: the variability of task-unrelated thoughts predicts error monitoring. Front Hum Neurosci, 7, 743. Amer, T., Anderson, J. A., Campbell, K. L., Hasher, L., & Grady, C. L. (2016). Age differences in the neural correlates of distraction regulation: A network interaction approach. Neuroimage, 139, 231-239. Amodio, D. M., & Frith, C. D. (2006). Meeting of minds: the medial frontal cortex and social cognition. Nat Rev Neurosci, 7(4), 268-277. Anderson, A. K. (2005). Affective influences on the attentional dynamics supporting awareness. Journal of experimental psychology. General, 134(2), 258-281. Anderson, A. K., & Phelps, E. A. (2001). Lesions of the human amygdala impair enhanced perception of emotionally salient events. Nature, 411(6835), 305-309. Anderson, J. S., Ferguson, M. A., Lopez-Larson, M., & Yurgelun-Todd, D. (2011). Connectivity gradients between the default mode and attention control networks. Brain connectivity, 1(2), 147-157. Andrews-Hanna, J. R. (2012). The brain's default network and its adaptive role in internal mentation. Neuroscientist, 18(3), 251-270. Andrews-Hanna, J. R., Reidler, J. S., Huang, C., & Buckner, R. L. (2010). Evidence for the default network's role in spontaneous cognition. J Neurophysiol, 104(1), 322-335. Andrews-Hanna, J. R., Reidler, J. S., Sepulcre, J., Poulin, R., & Buckner, R. L. (2010). Functional-anatomic fractionation of the brain's default network. Neuron, 65(4), 550-562. 130  Andrews-Hanna, J. R., Saxe, R., & Yarkoni, T. (2014). Contributions of episodic retrieval and mentalizing to autobiographical thought: Evidence from functional neuroimaging, resting-state connectivity, and fMRI meta-analyses. Neuroimage. Andrews-Hanna, J. R., Smallwood, J., & Spreng, R. N. (2014). The default network and self-generated thought: component processes, dynamic control, and clinical relevance. Ann N Y Acad Sci. Araujo, H. F., Kaplan, J., Damasio, H., & Damasio, A. (2015). Neural correlates of different self domains. Brain and Behavior. Arch, J. J., & Craske, M. G. (2006). Mechanisms of mindfulness: emotion regulation following a focused breathing induction. Behav Res Ther, 44(12), 1849-1858. Badre, D., & D'Esposito, M. (2009). Is the rostro-caudal axis of the frontal lobe hierarchical? Nat. Rev. Neurosci, 10(9), 659-669. Baird, B., Smallwood, J., Gorgolewski, K. J., & Margulies, D. S. (2013). Medial and lateral networks in anterior prefrontal cortex support metacognitive ability for memory and perception. J Neurosci, 33(42), 16657-16665. Baldauf, D., & Desimone, R. (2014). Neural mechanisms of object-based attention. Science, 344(6182), 424-427. Bar, K. J., de la Cruz, F., Schumann, A., Koehler, S., Sauer, H., Critchley, H., et al. (2016). Functional connectivity and network analysis of midbrain and brainstem nuclei. Neuroimage, 134, 53-63. Barcelo, F., Suwazono, S., & Knight, R. T. (2000). Prefrontal modulation of visual processing in humans. Nat Neurosci, 3(4), 399-403. Baron-Cohen, S., Wheelwright, S., Hill, J., Raste, Y., & Plumb, I. (2001). The "Reading the Mind in the Eyes" Test revised version: a study with normal adults, and adults with Asperger syndrome or high-functioning autism. J Child Psychol Psychiatry, 42(2), 241-251. Bartra, O., McGuire, J. T., & Kable, J. W. (2013). The valuation system: a coordinate-based meta-analysis of BOLD fMRI experiments examining neural correlates of subjective value. Neuroimage, 76, 412-427. Bassett, D. S., & Bullmore, E. (2006). Small-world brain networks. The Neuroscientist : a review journal bringing neurobiology, neurology and psychiatry, 12(6), 512-523. 131  Bassett, D. S., & Bullmore, E. T. (2016). Small-World Brain Networks Revisited. The Neuroscientist : a review journal bringing neurobiology, neurology and psychiatry. Bassett, D. S., Nelson, B. G., Mueller, B. A., Camchong, J., & Lim, K. O. (2012). Altered resting state complexity in schizophrenia. Neuroimage, 59(3), 2196-2207. Bassett, D. S., Wymbs, N. F., Porter, M. A., Mucha, P. J., Carlson, J. M., & Grafton, S. T. (2011). Dynamic reconfiguration of human brain networks during learning. Proceedings of the National Academy of Sciences of the United States of America, 108(18), 7641-7646. Bassett, D. S., Wymbs, N. F., Porter, M. A., Mucha, P. J., & Grafton, S. T. (2014). Cross-linked structure of network evolution. Chaos, 24(1), 013112. Beck, A. T. (1991). Cognitive therapy. A 30-year retrospective. Am Psychol, 46(4), 368-375. Behzadi, Y., Restom, K., Liau, J., & Liu, T. T. (2007). A component based noise correction method (CompCor) for BOLD and perfusion based fMRI. Neuroimage, 37(1), 90-101. Betzel, R. F., Fukushima, M., He, Y., Zuo, X. N., & Sporns, O. (2016). Dynamic fluctuations coincide with periods of high and low modularity in resting-state functional brain networks. Neuroimage, 127, 287-297. Bhatt, M. A., Lohrenz, T., Camerer, C. F., & Montague, P. R. (2010). Neural signatures of strategic types in a two-person bargaining game. Proc Natl Acad Sci U S A, 107(46), 19720-19725. Bichot, N. P., Heard, M. T., DeGennaro, E. M., & Desimone, R. (2015). A Source for Feature-Based Attention in the Prefrontal Cortex. Neuron, 88(4), 832-844. Binder, J. R., Desai, R. H., Graves, W. W., & Conant, L. L. (2009). Where is the semantic system? A critical review and meta-analysis of 120 functional neuroimaging studies. Cereb Cortex, 19(12), 2767-2796. Bisley, J. W., & Goldberg, M. E. (2003). Neuronal activity in the lateral intraparietal area and spatial attention. Science, 299(5603), 81-86. Biswal, B., Yetkin, F. Z., Haughton, V. M., & Hyde, J. S. (1995). Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn Reson Med, 34(4), 537-541. Blanke, O. (2012). Multisensory brain mechanisms of bodily self-consciousness. Nat Rev Neurosci, 13(8), 556-571. 132  Blondel, V. D., Guillaume, J.-L., Lambiotte, R., & Lefebvre, E. (2008). Fast unfolding of communities in large networks. Journal of statistical mechanics: theory and experiment, 2008(10), P10008. Brass, M., Derrfuss, J., Forstmann, B., & von Cramon, D. Y. (2005). The role of the inferior frontal junction area in cognitive control. Trends Cogn Sci, 9(7), 314-316. Braun, U., Schafer, A., Walter, H., Erk, S., Romanczuk-Seiferth, N., Haddad, L., et al. (2015). Dynamic reconfiguration of frontal brain networks during executive cognition in humans. Proc Natl Acad Sci U S A, 112(37), 11678-11683. Braver, T. S. (2012). The variable nature of cognitive control: a dual mechanisms framework. Trends Cogn Sci, 16(2), 106-113. Braver, T. S., & Bongiolatti, S. R. (2002). The role of frontopolar cortex in subgoal processing during working memory. NeuroImage, 15(3), 523-536. Brewer, J. A., Worhunsky, P. D., Gray, J. R., Tang, Y. Y., Weber, J., & Kober, H. (2011). Meditation experience is associated with differences in default mode network activity and connectivity. Proc Natl Acad Sci U S A, 108(50), 20254-20259. Buckner, R. L., Andrews-Hanna, J. R., & Schacter, D. L. (2008). The brain's default network: anatomy, function, and relevance to disease. Ann N Y Acad Sci, 1124, 1-38. Buckner, R. L., & Krienen, F. M. (2013). The evolution of distributed association networks in the human brain. Trends in cognitive sciences, 17(12), 648-665. Buckner, R. L., Krienen, F. M., & Yeo, B. T. (2013). Opportunities and limitations of intrinsic functional connectivity MRI. Nature neuroscience, 16(7), 832-837. Buckner, R. L., Sepulcre, J., Talukdar, T., Krienen, F. M., Liu, H., Hedden, T., et al. (2009). Cortical hubs revealed by intrinsic functional connectivity: mapping, assessment of stability, and relation to Alzheimer's disease. J Neurosci, 29(6), 1860-1873. Bullmore, E., & Sporns, O. (2009). Complex brain networks: graph theoretical analysis of structural and functional systems. Nat Rev Neurosci, 10(3), 186-198. Bunge, S. A., Kahn, I., Wallis, J. D., Miller, E. K., & Wagner, A. D. (2003). Neural circuits subserving the retrieval and maintenance of abstract rules. J Neurophysiol, 90(5), 3419-3428. Bunge, S. A., & Zelazo, P. D. (2006). A Brain-Based Account of the Development of Rule Use in Childhood. Curr Dir Psychol Sci, 15(3), 118-121  133  Burgess, P. W., Dumontheil, I., & Gilbert, S. J. (2007). The gateway hypothesis of rostral prefrontal cortex (area 10) function. Trends Cogn Sci, 11(7), 290-298. Buschman, T. J., & Kastner, S. (2015). From Behavior to Neural Dynamics: An Integrated Theory of Attention. Neuron, 88(1), 127-144. Buschman, T. J., & Miller, E. K. (2007). Top-down versus bottom-up control of attention in the prefrontal and posterior parietal cortices. Science, 315(5820), 1860-1862. Campbell-Sills, L., Barlow, D. H., Brown, T. A., & Hofmann, S. G. (2006). Effects of suppression and acceptance on emotional responses of individuals with anxiety and mood disorders. Behav Res Ther, 44(9), 1251-1263. Chai, X. J., Castañón, A. N., Öngür, D., & Whitfield-Gabrieli, S. (2012). Anticorrelations in resting state networks without global signal regression. Neuroimage, 59(2), 1420-1428. Chai, X. J., Ofen, N., Gabrieli, J. D., & Whitfield-Gabrieli, S. (2014). Selective development of anticorrelated networks in the intrinsic functional organization of the human brain. J Cogn Neurosci, 26(3), 501-513. Chang, C., & Glover, G. H. (2009). Effects of model-based physiological noise correction on default mode network anti-correlations and correlations. Neuroimage, 47(4), 1448-1459. Chang, C., & Glover, G. H. (2010). Time-frequency dynamics of resting-state brain connectivity measured with fMRI. Neuroimage, 50(1), 81-98. Chang, L. J., Yarkoni, T., Khaw, M. W., & Sanfey, A. G. (2013). Decoding the role of the insula in human cognition: functional parcellation and large-scale reverse inference. Cereb Cortex, 23(3), 739-749. Choi, E. Y., Yeo, B. T., & Buckner, R. L. (2012). The organization of the human striatum estimated by intrinsic functional connectivity. Journal of Neurophysiology, 108(8), 2242-2263. Christoff, K. (2012). Undirected thought: neural determinants and correlates. Brain Res, 1428, 51-59. Christoff, K., Cosmelli, D., Legrand, D., & Thompson, E. (2011). Specifying the self for cognitive neuroscience. Trends Cogn Sci, 15(3), 104-112. Christoff, K., & Gabrieli, J. D. E. (2000). The frontopolar cortex and human cognition: Evidence for a rostrocaudal hierarchical organization within the human prefrontal cortex. Psychobiology, 28(2), 168-186. 134  Christoff, K., Gordon, A. M., Smallwood, J., Smith, R., & Schooler, J. W. (2009). Experience sampling during fMRI reveals default network and executive system contributions to mind wandering. Proceedings of the National Academy of Sciences of the United States of America, 106(21), 8719-8724. Christoff, K., Irving, Z. C., Fox, K. C., Spreng, R. N., & Andrews-Hanna, J. R. (2016). Mind-wandering as spontaneous thought: a dynamic framework. Nature reviews. Neuroscience, 17(11), 718-731. Christoff, K., Irving, Z. C., Fox, K. C. R., Spreng, R. N., & Andrews-Hanna, J. R. (in press). Mind-wandering as spontaneous thought: A dynamic framework. Nat Rev Neurosci. Christoff, K., Keramatian, K., Gordon, A. M., Smith, R., & Madler, B. (2009). Prefrontal organization of cognitive control according to levels of abstraction. Brain research, 1286, 94-105. Christoff, K., Prabhakaran, V., Dorfman, J., Zhao, Z., Kroger, J. K., Holyoak, K. J., et al. (2001). Rostrolateral prefrontal cortex involvement in relational integration during reasoning. NeuroImage, 14(5), 1136-1149. Christoff, K., Ream, J. M., Geddes, L. P., & Gabrieli, J. D. (2003). Evaluating self-generated information: anterior prefrontal contributions to human cognition. Behavioral neuroscience, 117(6), 1161-1168. Chun, M. M., & Jiang, Y. (1998). Contextual cueing: implicit learning and memory of visual context guides spatial attention. Cogn Psychol, 36(1), 28-71. Chun, M. M., & Turk-Browne, N. B. (2007). Interactions between attention and memory. Curr Opin Neurobiol, 17(2), 177-184. Cocchi, L., Zalesky, A., Fornito, A., & Mattingley, J. B. (2013). Dynamic cooperation and competition between brain systems during cognitive control. Trends Cogn Sci, 17(10), 493-501. Cole, D. M., Smith, S. M., & Beckmann, C. F. (2010). Advances and pitfalls in the analysis and interpretation of resting-state FMRI data. Frontiers in systems neuroscience, 4, 8. Cole, M. W., Bagic, A., Kass, R., & Schneider, W. (2010). Prefrontal dynamics underlying rapid instructed task learning reverse with practice. J Neurosci, 30(42), 14245-14254. Cole, M. W., Bassett, D. S., Power, J. D., Braver, T. S., & Petersen, S. E. (2014). Intrinsic and task-evoked network architectures of the human brain. Neuron, 83(1), 238-251. 135  Cole, M. W., Ito, T., & Braver, T. S. (2015). The Behavioral Relevance of Task Information in Human Prefrontal Cortex. Cerebral cortex, bhv072. Cole, M. W., Laurent, P., & Stocco, A. (2013). Rapid instructed task learning: a new window into the human brain's unique capacity for flexible cognitive control. Cogn Affect Behav Neurosci, 13(1), 1-22. Cole, M. W., Repovs, G., & Anticevic, A. (2014). The frontoparietal control system: a central role in mental health. The Neuroscientist : a review journal bringing neurobiology, neurology and psychiatry, 20(6), 652-664. Cole, M. W., Reynolds, J. R., Power, J. D., Repovs, G., Anticevic, A., & Braver, T. S. (2013). Multi-task connectivity reveals flexible hubs for adaptive task control. Nat Neurosci, 16(9), 1348-1355. Cole, M. W., & Schneider, W. (2007). The cognitive control network: Integrated cortical regions with dissociable functions. Neuroimage, 37(1), 343-360. Cole, M. W., Yarkoni, T., Repovs, G., Anticevic, A., & Braver, T. S. (2012). Global connectivity of prefrontal cortex predicts cognitive control and intelligence. J Neurosci, 32(26), 8988-8999. Corbetta, M., Akbudak, E., Conturo, T. E., Snyder, A. Z., Ollinger, J. M., Drury, H. A., et al. (1998). A common network of functional areas for attention and eye movements. Neuron, 21(4), 761-773. Corbetta, M., Kincade, J. M., & Shulman, G. L. (2002). Neural systems for visual orienting and their relationships to spatial working memory. J Cogn Neurosci, 14(3), 508-523. Corbetta, M., & Shulman, G. L. (2002). Control of goal-directed and stimulus-driven attention in the brain. Nat Rev Neurosci, 3(3), 201-215. Craig, A. D. (2002). How do you feel? Interoception: the sense of the physiological condition of the body. Nat Rev Neurosci, 3(8), 655-666. Critchley, H. D., & Harrison, N. A. (2013). Visceral influences on brain and behavior. Neuron, 77(4), 624-638. Critchley, H. D., Wiens, S., Rotshtein, P., Ohman, A., & Dolan, R. J. (2004). Neural systems supporting interoceptive awareness. Nat Neurosci, 7(2), 189-195. Crittenden, B. M., Mitchell, D. J., & Duncan, J. (2015). Recruitment of the default mode network during a demanding act of executive control. Elife, 4, e06481. 136  Crittenden, B. M., Mitchell, D. J., & Duncan, J. (2016). Task Encoding across the Multiple Demand Cortex Is Consistent with a Frontoparietal and Cingulo-Opercular Dual Networks Distinction. The Journal of neuroscience : the official journal of the Society for Neuroscience, 36(23), 6147-6155. Crowe, D. A., Goodwin, S. J., Blackman, R. K., Sakellaridi, S., Sponheim, S. R., Macdonald, A. W., 3rd, et al. (2013). Prefrontal neurons transmit signals to parietal neurons that reflect executive control of cognition. Nat Neurosci, 16(10), 1484-1491. D'Argembeau, A., Collette, F., Van der Linden, M., Laureys, S., Del Fiore, G., Degueldre, C., et al. (2005). Self-referential reflective activity and its relationship with rest: a PET study. Neuroimage, 25(2), 616-624. D'Argembeau, A., Jedidi, H., Balteau, E., Bahri, M., Phillips, C., & Salmon, E. (2012). Valuing one's self: medial prefrontal involvement in epistemic and emotive investments in self-views. Cereb Cortex, 22(3), 659-667. D'Argembeau, A., Stawarczyk, D., Majerus, S., Collette, F., Van der Linden, M., Feyers, D., et al. (2010). The neural basis of personal goal processing when envisioning future events. J Cogn Neurosci, 22(8), 1701-1713. Damasio, A. R., & Carvalho, G. B. (2013). The nature of feelings: evolutionary and neurobiological origins. Nat Rev Neurosci, 14(2), 143-152. Davison, E. N., Schlesinger, K. J., Bassett, D. S., Lynall, M. E., Miller, M. B., Grafton, S. T., et al. (2015). Brain network adaptability across task states. PLoS computational biology, 11(1), e1004029. De Baene, W., Kuhn, S., & Brass, M. (2012). Challenging a decade of brain research on task switching: brain activation in the task-switching paradigm reflects adaptation rather than reconfiguration of task sets. Human brain mapping, 33(3), 639-651. De Havas, J. A., Parimal, S., Soon, C. S., & Chee, M. W. (2012). Sleep deprivation reduces default mode network connectivity and anti-correlation during rest and task performance. Neuroimage, 59(2), 1745-1751. De Martino, B., Fleming, S. M., Garrett, N., & Dolan, R. J. (2013). Confidence in value-based choice. Nat Neurosci, 16(1), 105-110. De Nooy, W., Mrvar, A., & Batagelj, V. (2011). Exploratory social network analysis with Pajek (Vol. 27): Cambridge University Press. 137  de Pasquale, F., Della Penna, S., Snyder, A. Z., Lewis, C., Mantini, D., Marzetti, L., et al. (2010). Temporal dynamics of spontaneous MEG activity in brain networks. Proceedings of the National Academy of Sciences of the United States of America, 107(13), 6040-6045. Deco, G., Jirsa, V. K., & McIntosh, A. R. (2011). Emerging concepts for the dynamical organization of resting-state activity in the brain. Nature reviews. Neuroscience, 12(1), 43-56. Delamillieure, P., Doucet, G., Mazoyer, B., Turbelin, M. R., Delcroix, N., Mellet, E., et al. (2010). The resting state questionnaire: An introspective questionnaire for evaluation of inner experience during the conscious resting state. Brain Res Bull, 81(6), 565-573. Denny, B. T., Kober, H., Wager, T. D., & Ochsner, K. N. (2012). A meta-analysis of functional neuroimaging studies of self- and other judgments reveals a spatial gradient for mentalizing in medial prefrontal cortex. J Cogn Neurosci, 24(8), 1742-1752. Desimone, R., & Duncan, J. (1995). Neural mechanisms of selective visual attention. Annu Rev Neurosci, 18, 193-222. Dixon, M. L., Andrews-Hanna, J. R., Spreng, R. N., Irving, Z. C., Mills, C., Girn, M., et al. (2017). Interactions between the default network and dorsal attention network vary across default subsystems, time, and cognitive states. Neuroimage, 147, 632-649. Dixon, M. L., & Christoff, K. (2012). The Decision to Engage Cognitive Control is Driven by Expected Reward-Value: Neural and Behavioral Evidence. PLoS One, 7(12), 1-12. Dixon, M. L., & Christoff, K. (2014). The lateral prefrontal cortex and complex value-based learning and decision making. Neurosci Biobehav Rev. Dixon, M. L., Fox, K. C. R., & Christoff, K. (2014a). Evidence for rostro-caudal functional organization in multiple brain areas related to goal-directed behavior. Brain Research. Dixon, M. L., Fox, K. C. R., & Christoff, K. (2014b). A framework for understanding the relationship between externally and internally directed cognition. Neuropsychologia, 62(321-330). Dixon, M. L., Girn, M., & Christoff, K. (2017). Hierarchical organization of frontoparietal control networks underlying goal-directed behavior. In M. Watanabe (Ed.), The Prefrontal Cortex as an Executive, Emotional, and Social Brain (pp. 133-148): Springer Japan. 138  Dixon, M. L., Thiruchselvam, R., Todd, R., & Christoff, K. (2017). Emotion and the Prefrontal Cortex: An Integrative Review. Psychological bulletin. Dosenbach, N. U., Fair, D. A., Miezin, F. M., Cohen, A. L., Wenger, K. K., Dosenbach, R. A., et al. (2007). Distinct brain networks for adaptive and stable task control in humans. Proc Natl Acad Sci U S A, 104(26), 11073-11078. Dosenbach, N. U., Nardos, B., Cohen, A. L., Fair, D. A., Power, J. D., Church, J. A., et al. (2010). Prediction of individual brain maturity using fMRI. Science, 329(5997), 1358-1361. Dosenbach, N. U., Visscher, K. M., Palmer, E. D., Miezin, F. M., Wenger, K. K., Kang, H. C., et al. (2006). A core system for the implementation of task sets. Neuron, 50(5), 799-812. Doucet, G., Naveau, M., Petit, L., Zago, L., Crivello, F., Jobard, G., et al. (2012). Patterns of hemodynamic low-frequency oscillations in the brain are modulated by the nature of free thought during rest. Neuroimage, 59(4), 3194-3200. Dunbar, R. I., & Shultz, S. (2007). Evolution in the social brain. Science, 317(5843), 1344-1347. Duncan, J. (2001). An adaptive coding model of neural function in prefrontal cortex. Nat Rev Neurosci, 2(11), 820-829. Duncan, J. (2010). The multiple-demand (MD) system of the primate brain: mental programs for intelligent behaviour. Trends Cogn Sci, 14(4), 172-179. Duncan, J. (2013). The structure of cognition: attentional episodes in mind and brain. Neuron, 80(1), 35-50. Duncan, J., Chylinski, D., Mitchell, D. J., & Bhandari, A. (2017). Complexity and compositionality in fluid intelligence. Proceedings of the National Academy of Sciences of the United States of America. Dwyer, D. B., Harrison, B. J., Yucel, M., Whittle, S., Zalesky, A., Pantelis, C., et al. (2014). Large-scale brain network dynamics supporting adolescent cognitive control. The Journal of neuroscience : the official journal of the Society for Neuroscience, 34(42), 14096-14107. Egner, T., & Hirsch, J. (2005). Cognitive control mechanisms resolve conflict through cortical amplification of task-relevant information. Nat Neurosci, 8(12), 1784-1790. Ellamil, M., Dobson, C., Beeman, M., & Christoff, K. (2012). Evaluative and generative modes of thought during the creative process. Neuroimage, 59(2), 1783-1794. 139  Ellamil, M., Fox, K. C., Dixon, M. L., Pritchard, S., Todd, R. M., Thompson, E., et al. (2016). Dynamics of neural recruitment surrounding the spontaneous arising of thoughts in experienced mindfulness practitioners. Neuroimage. Fair, D. A., Cohen, A. L., Dosenbach, N. U., Church, J. A., Miezin, F. M., Barch, D. M., et al. (2008). The maturing architecture of the brain's default network. Proc Natl Acad Sci U S A, 105(10), 4028-4032. Fair, D. A., Dosenbach, N. U., Church, J. A., Cohen, A. L., Brahmbhatt, S., Miezin, F. M., et al. (2007). Development of distinct control networks through segregation and integration. Proc Natl Acad Sci U S A, 104(33), 13507-13512. Farb, N. A., Anderson, A. K., Mayberg, H., Bean, J., McKeon, D., & Segal, Z. V. (2010). Minding one's emotions: mindfulness training alters the neural expression of sadness. Emotion, 10(1), 25-33. Farb, N. A., Segal, Z. V., & Anderson, A. K. (2012). Attentional Modulation of Primary Interoceptive and Exteroceptive Cortices. Cereb Cortex. Farb, N. A., Segal, Z. V., Mayberg, H., Bean, J., McKeon, D., Fatima, Z., et al. (2007). Attending to the present: mindfulness meditation reveals distinct neural modes of self-reference. Soc Cogn Affect Neurosci, 2(4), 313-322. Fazelpour, S., & Thompson, E. (2014). The Kantian brain: brain dynamics from a neurophenomenological perspective. Curr Opin Neurobiol, 31C, 223-229. Fleming, S. M., Weil, R. S., Nagy, Z., Dolan, R. J., & Rees, G. (2010). Relating introspective accuracy to individual differences in brain structure. Science, 329(5998), 1541-1543. Fletcher, P. C., Happe, F., Frith, U., Baker, S. C., Dolan, R. J., Frackowiak, R. S., et al. (1995). Other minds in the brain: a functional imaging study of "theory of mind" in story comprehension. Cognition, 57(2), 109-128. Fornito, A., Harrison, B. J., Zalesky, A., & Simons, J. S. (2012). Competitive and cooperative dynamics of large-scale brain functional networks supporting recollection. Proc Natl Acad Sci U S A, 109(31), 12788-12793. Fornito, A., Zalesky, A., & Bullmore, E. (2016). Fundamentals of brain network analysis: Academic Press. Fox, K., & Christoff, K. (2014). Metacognitive Facilitation of Spontaneous Thought Processes: When Metacognition Helps the Wandering Mind Find Its Way. In S. M. Fleming & C. D. 140  Frith (Eds.), The Cognitive Neuroscience of Metacognition (pp. 293-319). Berlin Heidelberg: Springer. Fox, K., Spreng, R. N., Ellamil, M., Andrews-Hanna, J. R., & Christoff, K. (2015). The wandering brain: Meta-analysis of functional neuroimaging studies of mind-wandering and related spontaneous thought processes. Neuroimage. Fox, K. C., Andrews-Hanna, J. R., & Christoff, K. (2016). The neurobiology of self-generated thought from cells to systems: Integrating evidence from lesion studies, human intracranial electrophysiology, neurochemistry, and neuroendocrinology. Neuroscience, 335, 134-150. Fox, M., Corbetta, M., Snyder, A. Z., Vincent, J. L., & Raichle, M. E. (2006). Spontaneous neuronal activity distinguishes human dorsal and ventral attention systems. Proc Natl Acad Sci U S A, 103(26), 10046-10051. Fox, M., & Raichle, M. E. (2007). Spontaneous fluctuations in brain activity observed with functional magnetic resonance imaging. Nat Rev Neurosci, 8(9), 700-711. Fox, M., Snyder, A. Z., Vincent, J. L., Corbetta, M., Van Essen, D. C., & Raichle, M. E. (2005). The human brain is intrinsically organized into dynamic, anticorrelated functional networks. Proc Natl Acad Sci U S A, 102(27), 9673-9678. Fox, M., Zhang, D., Snyder, A. Z., & Raichle, M. E. (2009). The global signal and observed anticorrelated resting state brain networks. J Neurophysiol, 101(6), 3270-3283. Fransson, P. (2005). Spontaneous low‐frequency BOLD signal fluctuations: An fMRI investigation of the resting‐state default mode of brain function hypothesis. Human brain mapping, 26(1), 15-29. Freedman, D. J., & Assad, J. A. (2006). Experience-dependent representation of visual categories in parietal cortex. Nature, 443(7107), 85-88. Gallagher, H. L., Happe, F., Brunswick, N., Fletcher, P. C., Frith, U., & Frith, C. D. (2000). Reading the mind in cartoons and stories: an fMRI study of 'theory of mind' in verbal and nonverbal tasks. Neuropsychologia, 38(1), 11-21. Gallagher, I. I. (2000). Philosophical conceptions of the self: implications for cognitive science. Trends Cogn Sci, 4(1), 14-21. 141  Gao, W., Gilmore, J. H., Shen, D., Smith, J. K., Zhu, H., & Lin, W. (2013). The synchronization within and interaction between the default and dorsal attention networks in early infancy. Cereb Cortex, 23(3), 594-603. Gao, W., & Lin, W. (2012). Frontal parietal control network regulates the anti-correlated default and dorsal attention networks. Hum Brain Mapp, 33(1), 192-202. Garfinkel, S. N., Zorab, E., Navaratnam, N., Engels, M., Mallorqui-Bague, N., Minati, L., et al. (2016). Anger in brain and body: the neural and physiological perturbation of decision-making by emotion. Social cognitive and affective neuroscience, 11(1), 150-158. Geerligs, L., Rubinov, M., Cam, C., & Henson, R. N. (2015). State and Trait Components of Functional Connectivity: Individual Differences Vary with Mental State. J Neurosci, 35(41), 13949-13961. Gerlach, K. D., Spreng, R. N., Madore, K. P., & Schacter, D. L. (2014). Future planning: Default network activity couples with frontoparietal control network and reward-processing regions during process and outcome simulations. Soc Cogn Affect Neurosci. Gilbert, S. J., Spengler, S., Simons, J. S., Steele, J. D., Lawrie, S. M., Frith, C. D., et al. (2006). Functional specialization within rostral prefrontal cortex (area 10): a meta-analysis. J Cogn Neurosci, 18(6), 932-948. Golland, Y., Bentin, S., Gelbard, H., Benjamini, Y., Heller, R., Nir, Y., et al. (2007). Extrinsic and intrinsic systems in the posterior cortex of the human brain revealed during natural sensory stimulation. Cereb Cortex, 17(4), 766-777. Golland, Y., Golland, P., Bentin, S., & Malach, R. (2008). Data-driven clustering reveals a fundamental subdivision of the human cortex into two global systems. Neuropsychologia, 46(2), 540-553. Gollo, L. L., Zalesky, A., Hutchison, R. M., van den Heuvel, M., & Breakspear, M. (2015). Dwelling quietly in the rich club: brain network determinants of slow cortical fluctuations. Philos Trans R Soc Lond B Biol Sci, 370(1668). Gollwitzer, P. M. (1999). Implementation intentions: strong effects of simple plans. American Psychologist, 54(7), 493-503. Gonzalez-Castillo, J., Hoy, C. W., Handwerker, D. A., Robinson, M. E., Buchanan, L. C., Saad, Z. S., et al. (2015). Tracking ongoing cognition in individuals using brief, whole-brain functional connectivity patterns. Proc Natl Acad Sci U S A, 112(28), 8762-8767. 142  Gordon, E. M., Laumann, T. O., Adeyemo, B., Huckins, J. F., Kelley, W. M., & Petersen, S. E. (2014). Generation and Evaluation of a Cortical Area Parcellation from Resting-State Correlations. Cereb Cortex. Gorgolewski, K. J., Lurie, D., Urchs, S., Kipping, J. A., Craddock, R. C., Milham, M. P., et al. (2014). A correspondence between individual differences in the brain's intrinsic functional architecture and the content and form of self-generated thoughts. PLoS One, 9(5), e97176. Gottlieb, J. P., Kusunoki, M., & Goldberg, M. E. (1998). The representation of visual salience in monkey parietal cortex. Nature, 391(6666), 481-484. Greicius, M., Krasnow, B., Reiss, A. L., & Menon, V. (2003). Functional connectivity in the resting brain: a network analysis of the default mode hypothesis. Proc Natl Acad Sci U S A, 100(1), 253-258. Gross, J. J. (1998). The emerging field of emotion regulation: an integrative review. Review of general psychology, 2(3), 271. Guimera, R., & Nunes Amaral, L. A. (2005). Functional cartography of complex metabolic networks. Nature, 433(7028), 895-900. Gusnard, D. A., Akbudak, E., Shulman, G. L., & Raichle, M. E. (2001). Medial prefrontal cortex and self-referential mental activity: relation to a default mode of brain function. Proc Natl Acad Sci U S A, 98(7), 4259-4264. Gusnard, D. A., & Raichle, M. E. (2001). Searching for a baseline: functional imaging and the resting human brain. Nat Rev Neurosci, 2(10), 685-694. Hagmann, P., Cammoun, L., Gigandet, X., Meuli, R., Honey, C. J., Wedeen, V. J., et al. (2008). Mapping the structural core of human cerebral cortex. PLoS Biol, 6(7), e159. Hampshire, A., Thompson, R., Duncan, J., & Owen, A. M. (2009). Selective tuning of the right inferior frontal gyrus during target detection. Cogn Affect Behav Neurosci, 9(1), 103-112. Hare, T. A., Malmaud, J., & Rangel, A. (2011). Focusing Attention on the Health Aspects of Foods Changes Value Signals in vmPFC and Improves Dietary Choice. J Neurosci, 31(30), 11077-11087. Hasenkamp, W., & Barsalou, L. W. (2012). Effects of meditation experience on functional connectivity of distributed brain networks. Frontiers in human neuroscience, 6, 38. 143  Hassabis, D., Kumaran, D., & Maguire, E. A. (2007). Using imagination to understand the neural basis of episodic memory. J Neurosci, 27(52), 14365-14374. Hassabis, D., Spreng, R. N., Rusu, A. A., Robbins, C. A., Mar, R. A., & Schacter, D. L. (2013). Imagine All the People: How the Brain Creates and Uses Personality Models to Predict Behavior. Cereb Cortex. Hasselmo, M. E., & McGaughy, J. (2004). High acetylcholine levels set circuit dynamics for attention and encoding and low acetylcholine levels set dynamics for consolidation. Progress in brain research, 145, 207-231. Hayes, S. C., Luoma, J. B., Bond, F. W., Masuda, A., & Lillis, J. (2006). Acceptance and commitment therapy: model, processes and outcomes. Behav Res Ther, 44(1), 1-25. Hermundstad, A. M., Brown, K. S., Bassett, D. S., Aminoff, E. M., Frithsen, A., Johnson, A., et al. (2014). Structurally-constrained relationships between cognitive states in the human brain. PLoS Comput Biol, 10(5), e1003591. Hill, J., Inder, T., Neil, J., Dierker, D., Harwell, J., & Van Essen, D. (2010). Similar patterns of cortical expansion during human development and evolution. Proceedings of the National Academy of Sciences of the United States of America, 107(29), 13135-13140. Holmes, A. J., Hollinshead, M. O., O'Keefe, T. M., Petrov, V. I., Fariello, G. R., Wald, L. L., et al. (2015). Brain Genomics Superstruct Project initial data release with structural, functional, and behavioral measures. Scientific data, 2, 150031. Honey, C. J., Sporns, O., Cammoun, L., Gigandet, X., Thiran, J. P., Meuli, R., et al. (2009). Predicting human resting-state functional connectivity from structural connectivity. Proc Natl Acad Sci U S A, 106(6), 2035-2040. Hopfinger, J. B., Buonocore, M. H., & Mangun, G. R. (2000). The neural mechanisms of top-down attentional control. Nat Neurosci, 3(3), 284-291. Huijbers, W., Pennartz, C. M., Cabeza, R., & Daselaar, S. M. (2011). The hippocampus is coupled with the default network during memory retrieval but not during memory encoding. PLoS One, 6(4), e17463. Hutchison, R. M., Womelsdorf, T., Allen, E. A., Bandettini, P. A., Calhoun, V. D., Corbetta, M., et al. (2013). Dynamic functional connectivity: promise, issues, and interpretations. Neuroimage, 80, 360-378. 144  Hutchison, R. M., Womelsdorf, T., Gati, J. S., Everling, S., & Menon, R. S. (2013). Resting-state networks show dynamic functional connectivity in awake humans and anesthetized macaques. Hum Brain Mapp, 34(9), 2154-2177. Itti, L., & Koch, C. (2001). Computational modelling of visual attention. Nature reviews. Neuroscience, 2(3), 194-203. Jack, A. I., Dawson, A. J., Begany, K. L., Leckie, R. L., Barry, K. P., Ciccia, A. H., et al. (2012). fMRI reveals reciprocal inhibition between social and physical cognitive domains. Neuroimage, 66C, 385-401. Jang, J. H., Jung, W. H., Kang, D. H., Byun, M. S., Kwon, S. J., Choi, C. H., et al. (2011). Increased default mode network connectivity associated with meditation. Neuroscience letters, 487(3), 358-362. Jao, T., Li, C. W., Vertes, P. E., Wu, C. W., Achard, S., Hsieh, C. H., et al. (2016). Large-Scale Functional Brain Network Reorganization During Taoist Meditation. Brain connectivity, 6(1), 9-24. Jimura, K., Chushak, M. S., & Braver, T. S. (2013). Impulsivity and self-control during intertemporal decision making linked to the neural dynamics of reward value representation. J Neurosci, 33(1), 344-357. Jimura, K., Locke, H. S., & Braver, T. S. (2010). Prefrontal cortex mediation of cognitive enhancement in rewarding motivational contexts. Proc Natl Acad Sci U S A, 107(19), 8871-8876. Josipovic, Z., Dinstein, I., Weber, J., & Heeger, D. J. (2011). Influence of meditation on anti-correlated networks in the brain. Frontiers in human neuroscience, 5. Josipovic, Z., Dinstein, I., Weber, J., & Heeger, D. J. (2012). Influence of meditation on anti-correlated networks in the brain. Frontiers in human neuroscience, 5, 183. Kabat-Zinn, J. (2009). Wherever you go, there you are. New York, NY: MJF Books. Kam, J. W., & Handy, T. C. (2014). Differential recruitment of executive resources during mind wandering. Conscious Cogn, 26C, 51-63. Kamada, T., & Kawai, S. (1989). An algorithm for drawing general undirected graphs. Information processing letters, 31(1), 7-15. 145  Kastner, S., De Weerd, P., Desimone, R., & Ungerleider, L. G. (1998). Mechanisms of directed attention in the human extrastriate cortex as revealed by functional MRI. Science, 282(5386), 108-111. Kastner, S., Pinsk, M. A., De Weerd, P., Desimone, R., & Ungerleider, L. G. (1999). Increased activity in human visual cortex during directed attention in the absence of visual stimulation. Neuron, 22(4), 751-761. Keller, J. B., Hedden, T., Thompson, T. W., Anteraper, S. A., Gabrieli, J. D., & Whitfield-Gabrieli, S. (2015). Resting-state anticorrelations between medial and lateral prefrontal cortex: association with working memory, aging, and individual differences. Cortex; a journal devoted to the study of the nervous system and behavior, 64, 271-280. Kelly, A. M., Uddin, L. Q., Biswal, B. B., Castellanos, F. X., & Milham, M. P. (2008). Competition between functional brain networks mediates behavioral variability. Neuroimage, 39(1), 527-537. Kleckner, I., Zhang, J., Touroutoglou, A., Chanes, L., Xia, C., Simmons, W., Quigley, K., Dickerson, B., & Feldman Barrett, L. . (2017). Evidence for a large-scale brain system supporting allostasis and interoception in humans Nature Human Behaviour. Koechlin, E. (2011). Frontal pole function: what is specifically human? Trends Cogn Sci, 15(6), 241; author reply 243. Koechlin, E., Basso, G., Pietrini, P., Panzer, S., & Grafman, J. (1999). The role of the anterior prefrontal cortex in human cognition. Nature, 399(6732), 148-151. Koechlin, E., Ody, C., & Kouneiher, F. (2003). The architecture of cognitive control in the human prefrontal cortex. Science, 302(5648), 1181-1185. Koechlin, E., & Summerfield, C. (2007). An information theoretical approach to prefrontal executive function. Trends Cogn Sci, 11(6), 229-235. Konishi, M., McLaren, D. G., Engen, H., & Smallwood, J. (2015). Shaped by the Past: The Default Mode Network Supports Cognition that Is Independent of Immediate Perceptual Input. PLoS One, 10(6), e0132209. Krajbich, I., Armel, C., & Rangel, A. (2010). Visual fixations and the computation and comparison of value in simple choice. Nature neuroscience, 13(10), 1292-1298. Krieger-Redwood, K., Jefferies, E., Karapanagiotidis, T., Seymour, R., Nunes, A., Ang, J. W., et al. (2016). Down but not out in posterior cingulate cortex: Deactivation yet functional 146  coupling with prefrontal cortex during demanding semantic cognition. Neuroimage, 141, 366-377. Krienen, F. M., Yeo, B. T., & Buckner, R. L. (2014). Reconfigurable task-dependent functional coupling modes cluster around a core functional architecture. Philos Trans R Soc Lond B Biol Sci, 369(1653). Kucyi, A., Hove, M. J., Esterman, M., Hutchison, R. M., & Valera, E. M. (2016). Dynamic Brain Network Correlates of Spontaneous Fluctuations in Attention. Cerebral cortex. Lane, R. D., Fink, G. R., Chau, P. M., & Dolan, R. J. (1997). Neural activation during selective attention to subjective emotional responses. Neuroreport, 8(18), 3969-3972. Latora, V., & Marchiori, M. (2001). Efficient behavior of small-world networks. Physical review letters, 87(19), 198701. Laumann, T. O., Snyder, A. Z., Mitra, A., Gordon, E. M., Gratton, C., Adeyemo, B., et al. (2016). On the Stability of BOLD fMRI Correlations. Cerebral cortex. Lee, K. H., & Siegle, G. J. (2012). Common and distinct brain networks underlying explicit emotional evaluation: a meta-analytic study. Soc Cogn Affect Neurosci, 7(5), 521-534. Lee, M. H., Hacker, C. D., Snyder, A. Z., Corbetta, M., Zhang, D., Leuthardt, E. C., et al. (2012). Clustering of resting state networks. PLoS One, 7(7), e40370. Leech, R., Braga, R., & Sharp, D. J. (2012). Echoes of the brain within the posterior cingulate cortex. J Neurosci, 32(1), 215-222. Leonardi, N., & Van De Ville, D. (2015). On spurious and real fluctuations of dynamic functional connectivity during rest. Neuroimage, 104, 430-436. Liegeois, R., Ziegler, E., Phillips, C., Geurts, P., Gomez, F., Bahri, M. A., et al. (2015). Cerebral functional connectivity periodically (de)synchronizes with anatomical constraints. Brain structure & function. Lindquist, K. A., Satpute, A. B., Wager, T. D., Weber, J., & Barrett, L. F. (2016). The Brain Basis of Positive and Negative Affect: Evidence from a Meta-Analysis of the Human Neuroimaging Literature. Cerebral cortex, 26(5), 1910-1922. Loftus, G. R., & Masson, E. J. M. (1994). Using confidence intervals in within–subject designs. Psychon Bull Rev, 1, 476–490. 147  MacDonald, A. W., 3rd, Cohen, J. D., Stenger, V. A., & Carter, C. S. (2000). Dissociating the role of the dorsolateral prefrontal and anterior cingulate cortex in cognitive control. Science, 288(5472), 1835-1838. Mantini, D., Perrucci, M. G., Del Gratta, C., Romani, G. L., & Corbetta, M. (2007). Electrophysiological signatures of resting state networks in the human brain. Proceedings of the National Academy of Sciences of the United States of America, 104(32), 13170-13175. Mar, R. A. (2011). The neural bases of social cognition and story comprehension. Annual Review of Psychology, 62, 103-134. Margulies, D. S., Ghosh, S. S., Goulas, A., Falkiewicz, M., Huntenburg, J. M., Langs, G., et al. (2016). Situating the default-mode network along a principal gradient of macroscale cortical organization. Proceedings of the National Academy of Sciences of the United States of America, 113(44), 12574-12579. Margulies, D. S., Vincent, J. L., Kelly, C., Lohmann, G., Uddin, L. Q., Biswal, B. B., et al. (2009). Precuneus shares intrinsic functional architecture in humans and monkeys. Proc Natl Acad Sci U S A, 106(47), 20069-20074. Marois, R., & Ivanoff, J. (2005). Capacity limits of information processing in the brain. Trends Cogn Sci, 9(6), 296-305. Mazoyer, B., Zago, L., Mellet, E., Bricogne, S., Etard, O., Houde, O., et al. (2001). Cortical networks for working memory and executive functions sustain the conscious resting state in man. Brain Res Bull, 54(3), 287-298. McCaig, R. G., Dixon, M., Keramatian, K., Liu, I., & Christoff, K. (2011). Improved modulation of rostrolateral prefrontal cortex using real-time fMRI training and meta-cognitive awareness. Neuroimage, 55(3), 1298-1305. McCurdy, L. Y., Maniscalco, B., Metcalfe, J., Liu, K. Y., de Lange, F. P., & Lau, H. (2013). Anatomical coupling between distinct metacognitive systems for memory and visual perception. J Neurosci, 33(5), 1897-1906. Medaglia, J. D., Lynall, M. E., & Bassett, D. S. (2015). Cognitive network neuroscience. Journal of Cognitive Neuroscience, 27(8), 1471-1491. 148  Mennes, M., Kelly, C., Colcombe, S., Castellanos, F. X., & Milham, M. P. (2013). The extrinsic and intrinsic functional architectures of the human brain are not equivalent. Cereb Cortex, 23(1), 223-229. Menon, V., & Uddin, L. Q. (2010). Saliency, switching, attention and control: a network model of insula function. Brain Struct Funct, 214(5-6), 655-667. Mesulam, M. (1998). From sensation to cognition. Brain, 121 ( Pt 6), 1013-1052. Meyer, M. L., Spunt, R. P., Berkman, E. T., Taylor, S. E., & Lieberman, M. D. (2012). Evidence for social working memory from a parametric functional MRI study. Proc Natl Acad Sci U S A, 109(6), 1883-1888. Milazzo, A. C., Ng, B., Jiang, H., Shirer, W., Varoquaux, G., Poline, J. B., et al. (2014). Identification of Mood-Relevant Brain Connections Using a Continuous, Subject-Driven Rumination Paradigm. Cereb Cortex. Miller, E. K., & Buschman, T. J. (2013). Cortical circuits for the control of attention. Current opinion in neurobiology, 23(2), 216-222. Miller, E. K., & Cohen, J. D. (2001). An integrative theory of prefrontal cortex function. Annu Rev Neurosci, 24, 167-202. Mitchell, D. J., Bell, A. H., Buckley, M. J., Mitchell, A. S., Sallet, J., & Duncan, J. (2016). A Putative Multiple-Demand System in the Macaque Brain. The Journal of neuroscience : the official journal of the Society for Neuroscience, 36(33), 8574-8585. Mitchell, J. P., Macrae, C. N., & Banaji, M. R. (2006). Dissociable medial prefrontal contributions to judgments of similar and dissimilar others. Neuron, 50(4), 655-663. Mitra, A., Snyder, A. Z., Blazey, T., & Raichle, M. E. (2015). Lag threads organize the brain's intrinsic activity. Proceedings of the National Academy of Sciences of the United States of America, 112(17), E2235-2244. Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, A. H., Howerter, A., & Wager, T. D. (2000). The unity and diversity of executive functions and their contributions to complex "Frontal Lobe" tasks: a latent variable analysis. Cognitive psychology, 41(1), 49-100. Moffitt, T. E., Arseneault, L., Belsky, D., Dickson, N., Hancox, R. J., Harrington, H., et al. (2011). A gradient of childhood self-control predicts health, wealth, and public safety. Proc Natl Acad Sci U S A, 108(7), 2693-2698. 149  Moore, T., & Armstrong, K. M. (2003). Selective gating of visual signals by microstimulation of frontal cortex. Nature, 421(6921), 370-373. Moran, J., & Desimone, R. (1985). Selective attention gates visual processing in the extrastriate cortex. Science, 229(4715), 782-784. Morcom, A. M., & Fletcher, P. C. (2007). Does the brain have a baseline? Why we should be resisting a rest. Neuroimage, 37(4), 1073-1082. Morecraft, R. J., Geula, C., & Mesulam, M. M. (1993). Architecture of Connectivity within a Cingulo-Fronto-Parietal Neurocognitive Network for Directed Attention. Archives of Neurology, 50(3), 279-284. Mueller, S., Wang, D., Fox, M. D., Yeo, B. T., Sepulcre, J., Sabuncu, M. R., et al. (2013). Individual variability in functional connectivity architecture of the human brain. Neuron, 77(3), 586-595. Murphy, A. C., Gu, S., Khambhati, A. N., Wymbs, N. F., Grafton, S. T., Satterthwaite, T. D., et al. (2016). Explicitly Linking Regional Activation and Function Connectivity: Community Structure of Weighted Networks with Continuous Annotation. arXiv preprint arXiv:1611.07962. Murphy, K., Birn, R. M., Handwerker, D. A., Jones, T. B., & Bandettini, P. A. (2009). The impact of global signal regression on resting state correlations: are anti-correlated networks introduced? Neuroimage, 44(3), 893-905. Murray, R. J., Schaer, M., & Debbane, M. (2012). Degrees of separation: a quantitative neuroimaging meta-analysis investigating self-specificity and shared neural activation between self- and other-reflection. Neurosci Biobehav Rev, 36(3), 1043-1059. Nee, D. E., & D'Esposito, M. (2016). The hierarchical organization of the lateral prefrontal cortex. Elife, 5. Newman, M. E. (2004). Fast algorithm for detecting community structure in networks. Physical review. E, Statistical, nonlinear, and soft matter physics, 69(6 Pt 2), 066133. Nolen-Hoeksema, S. (2000). The role of rumination in depressive disorders and mixed anxiety/depressive symptoms. J Abnorm Psychol, 109(3), 504-511. O'Reilly, R. C., Herd, S. A., & Pauli, W. M. (2010). Computational models of cognitive control. Current opinion in neurobiology, 20(2), 257-261. 150  Ochsner, K. N., Beer, J. S., Robertson, E. R., Cooper, J. C., Gabrieli, J. D., Kihsltrom, J. F., et al. (2005). The neural correlates of direct and reflected self-knowledge. Neuroimage, 28(4), 797-814. Onnela, J. P., Saramaki, J., Kertesz, J., & Kaski, K. (2005). Intensity and coherence of motifs in weighted complex networks. Physical review. E, Statistical, nonlinear, and soft matter physics, 71(6 Pt 2), 065103. Opsahl, T., Colizza, V., Panzarasa, P., & Ramasco, J. J. (2008). Prominence and control: the weighted rich-club effect. Physical review letters, 101(16), 168702. Pagnoni, G. (2012). Dynamical properties of BOLD activity from the ventral posteromedial cortex associated with meditation and attentional skills. The Journal of neuroscience : the official journal of the Society for Neuroscience, 32(15), 5242-5249. Parro, C., Dixon, M. L., & Christoff, K. (2017). The Neural Basis of Motivational Influences on Cognitive Control: An ALE Meta-Analysis. bioRxiv, 113126. Passingham, R. E., Stephan, K. E., & Kotter, R. (2002). The anatomical basis of functional localization in the cortex. Nat Rev Neurosci, 3(8), 606-616. Passingham, R. E., & Wise, S. P. (2012). The neurobiology of the prefrontal cortex : anatomy, evolution, and the origin of insight (1st ed.). Oxford: Oxford University Press. Pessoa, L. (2005). To what extent are emotional visual stimuli processed without attention and awareness? Curr Opin Neurobiol, 15(2), 188-196. Pessoa, L. (2014). Understanding brain networks and brain organization. Physics of life reviews, 11(3), 400-435. Petersen, S. E., & Sporns, O. (2015). Brain Networks and Cognitive Architectures. Neuron, 88(1), 207-219. Petrides, M. (2005). Lateral prefrontal cortex: architectonic and functional organization. Philos Trans R Soc Lond B Biol Sci, 360(1456), 781-795. Poldrack, R. A., Laumann, T. O., Koyejo, O., Gregory, B., Hover, A., Chen, M. Y., et al. (2015). Long-term neural and physiological phenotyping of a single human. Nature communications, 6, 8885. Poppenk, J., Moscovitch, M., McIntosh, A. R., Ozcelik, E., & Craik, F. I. (2010). Encoding the future: successful processing of intentions engages predictive brain networks. Neuroimage, 49(1), 905-913. 151  Posner, M. I. (2012). Imaging attention networks. Neuroimage, 61(2), 450-456. Posner, M. I., & Dehaene, S. (1994). Attentional networks. Trends Neurosci, 17(2), 75-79. Posner, M. I., & DiGirolamo, G. J. (1998). Conflict, target detection and cognitive control. The attentive brain, 401-423. Power, J. D., Cohen, A. L., Nelson, S. M., Wig, G. S., Barnes, K. A., Church, J. A., et al. (2011). Functional network organization of the human brain. Neuron, 72(4), 665-678. Power, J. D., Schlaggar, B. L., Lessov-Schlaggar, C. N., & Petersen, S. E. (2013). Evidence for hubs in human functional brain networks. Neuron, 79(4), 798-813. Ptak, R. (2012). The frontoparietal attention network of the human brain: action, saliency, and a priority map of the environment. The Neuroscientist : a review journal bringing neurobiology, neurology and psychiatry, 18(5), 502-515. Raichle, M. E. (2009). A paradigm shift in functional brain imaging. J Neurosci, 29(41), 12729-12734. Raichle, M. E. (2015). The brain's default mode network. Annual review of neuroscience, 38, 433-447. Raichle, M. E., MacLeod, A. M., Snyder, A. Z., Powers, W. J., Gusnard, D. A., & Shulman, G. L. (2001). A default mode of brain function. Proc Natl Acad Sci U S A, 98(2), 676-682. Reichardt, J., & Bornholdt, S. (2006). Statistical mechanics of community detection. Physical Review E, 74(1), 016110. Reineberg, A. E., Andrews-Hanna, J. R., Depue, B. E., Friedman, N. P., & Banich, M. T. (2015). Resting-state networks predict individual differences in common and specific aspects of executive function. Neuroimage, 104, 69-78. Repovs, G., & Barch, D. M. (2012). Working memory related brain network connectivity in individuals with schizophrenia and their siblings. Frontiers in human neuroscience, 6, 137. Rubinov, M., & Sporns, O. (2010). Complex network measures of brain connectivity: uses and interpretations. Neuroimage, 52(3), 1059-1069. Saad, Z. S., Gotts, S. J., Murphy, K., Chen, G., Jo, H. J., Martin, A., et al. (2012). Trouble at rest: how correlation patterns and group differences become distorted after global signal regression. Brain connectivity, 2(1), 25-32. Sakai, K. (2008). Task set and prefrontal cortex. Annu Rev Neurosci, 31, 219-245. 152  Saldanha, A. J. (2004). Java Treeview—extensible visualization of microarray data. Bioinformatics, 20(17), 3246-3248. Schacter, D. L., Addis, D. R., & Buckner, R. L. (2007). Remembering the past to imagine the future: the prospective brain. Nat Rev Neurosci, 8(9), 657-661. Schmitz, T. W., & Johnson, S. C. (2007). Relevance to self: A brief review and framework of neural systems underlying appraisal. Neurosci Biobehav Rev, 31(4), 585-596. Schooler, J. W., Smallwood, J., Christoff, K., Handy, T. C., Reichle, E. D., & Sayette, M. A. (2011). Meta-awareness, perceptual decoupling and the wandering mind. Trends Cogn Sci, 15(7), 319-326. Schultz, D. H., & Cole, M. W. (2016). Higher Intelligence Is Associated with Less Task-Related Brain Network Reconfiguration. The Journal of neuroscience : the official journal of the Society for Neuroscience, 36(33), 8551-8561. Seeley, W. W., Menon, V., Schatzberg, A. F., Keller, J., Glover, G. H., Kenna, H., et al. (2007). Dissociable intrinsic connectivity networks for salience processing and executive control. J Neurosci, 27(9), 2349-2356. Segal, Z. V., Williams, J. M. G., & Teasdale, J. D. (2013). Mindfulness-based cognitive therapy for depression (2nd ed.). New York: Guilford Press. Serences, J. T., & Yantis, S. (2006). Selective visual attention and perceptual coherence. Trends Cogn Sci, 10(1), 38-45. Shine, J. M., Bissett, P. G., Bell, P. T., Koyejo, O., Balsters, J. H., Gorgolewski, K. J., et al. (2016). The Dynamics of Functional Brain Networks: Integrated Network States during Cognitive Task Performance. Neuron. Shine, J. M., Koyejo, O., & Poldrack, R. A. (2016). Temporal metastates are associated with differential patterns of time-resolved connectivity, network topology, and attention. Proceedings of the National Academy of Sciences of the United States of America, 113(35), 9888-9891. Shirer, W. R., Ryali, S., Rykhlevskaia, E., Menon, V., & Greicius, M. D. (2012). Decoding subject-driven cognitive states with whole-brain connectivity patterns. Cereb Cortex, 22(1), 158-165. 153  Shulman, G. L., Fiez, J. A., Corbetta, M., Buckner, R. L., Miezin, F. M., Raichle, M. E., et al. (1997). Common blood flow changes across visual tasks: II. Decreases in cerebral cortex. Journal of Cognitive Neuroscience, 9(5), 648-663. Simons, J. S., & Spiers, H. J. (2003). Prefrontal and medial temporal lobe interactions in long-term memory. Nat Rev Neurosci, 4(8), 637-648. Simony, E., Honey, C. J., Chen, J., Lositsky, O., Yeshurun, Y., Wiesel, A., et al. (2016). Dynamic reconfiguration of the default mode network during narrative comprehension. Nature communications, 7, 12141. Skocik, M., Collins, J., Callahan-Flintoft, C., Bowman, H., & Wyble, B. (2016). I TRIED A BUNCH OF THINGS: THE DANGERS OF UNEXPECTED OVERFITTING IN CLASSIFICATION. bioRxiv, 078816. Smallwood, J., Brown, K., Baird, B., & Schooler, J. W. (2012). Cooperation between the default mode network and the frontal-parietal network in the production of an internal train of thought. Brain Res, 1428, 60-70. Smallwood, J., McSpadden, M., & Schooler, J. W. (2008). When attention matters: the curious incident of the wandering mind. Mem Cognit, 36(6), 1144-1150. Smallwood, J., & Schooler, J. W. (2006). The restless mind. Psychol Bull, 132(6), 946-958. Smith, S. M., Fox, P. T., Miller, K. L., Glahn, D. C., Fox, P. M., Mackay, C. E., et al. (2009). Correspondence of the brain's functional architecture during activation and rest. Proc Natl Acad Sci U S A, 106(31), 13040-13045. Smith, S. M., Vidaurre, D., Beckmann, C. F., Glasser, M. F., Jenkinson, M., Miller, K. L., et al. (2013). Functional connectomics from resting-state fMRI. Trends Cogn Sci, 17(12), 666-682. Sonuga-Barke, E. J., & Castellanos, F. X. (2007). Spontaneous attentional fluctuations in impaired states and pathological conditions: a neurobiological hypothesis. Neuroscience & Biobehavioral Reviews, 31(7), 977-986. Sporns, O., Chialvo, D. R., Kaiser, M., & Hilgetag, C. C. (2004). Organization, development and function of complex brain networks. Trends Cogn Sci, 8(9), 418-425. Spreng, R. N. (2012). The fallacy of a "task-negative" network. Front Psychol, 3, 145. 154  Spreng, R. N., DuPre, E., Selarka, D., Garcia, J., Gojkovic, S., Mildner, J., et al. (2014). Goal-congruent default network activity facilitates cognitive control. J Neurosci, 34(42), 14108-14114. Spreng, R. N., Mar, R. A., & Kim, A. S. (2009). The common neural basis of autobiographical memory, prospection, navigation, theory of mind, and the default mode: a quantitative meta-analysis. J Cogn Neurosci, 21(3), 489-510. Spreng, R. N., Sepulcre, J., Turner, G. R., Stevens, W. D., & Schacter, D. L. (2013). Intrinsic architecture underlying the relations among the default, dorsal attention, and frontoparietal control networks of the human brain. J Cogn Neurosci, 25(1), 74-86. Spreng, R. N., Stevens, W. D., Chamberlain, J. P., Gilmore, A. W., & Schacter, D. L. (2010). Default network activity, coupled with the frontoparietal control network, supports goal-directed cognition. Neuroimage, 53(1), 303-317. Spreng, R. N., Stevens, W. D., Viviano, J. D., & Schacter, D. L. (2016). Attenuated anticorrelation between the default and dorsal attention networks with aging: evidence from task and rest. Neurobiology of Aging, 45, 149-160. Sridharan, D., Levitin, D. J., & Menon, V. (2008). A critical role for the right fronto-insular cortex in switching between central-executive and default-mode networks. Proc Natl Acad Sci U S A, 105(34), 12569-12574. Stokes, M. G., Kusunoki, M., Sigala, N., Nili, H., Gaffan, D., & Duncan, J. (2013). Dynamic coding for cognitive control in prefrontal cortex. Neuron, 78(2), 364-375. Stuss, D. T., & Knight, R. T. (2002). Principles of frontal lobe function. Oxford ; New York: Oxford University Press. Suddendorf, T., & Corballis, M. C. (2007). The evolution of foresight: What is mental time travel, and is it unique to humans? The Behavioral and brain sciences, 30(3), 299-313; discussion 313-251. Summerfield, J. J., Lepsien, J., Gitelman, D. R., Mesulam, M. M., & Nobre, A. C. (2006). Orienting attention based on long-term memory experience. Neuron, 49(6), 905-916. Svoboda, E., McKinnon, M. C., & Levine, B. (2006). The functional neuroanatomy of autobiographical memory: a meta-analysis. Neuropsychologia, 44(12), 2189-2208. 155  Tagliazucchi, E., Roseman, L., Kaelen, M., Orban, C., Muthukumaraswamy, S. D., Murphy, K., et al. (2016). Increased Global Functional Connectivity Correlates with LSD-Induced Ego Dissolution. Current biology : CB, 26(8), 1043-1050. Tavares, P., Lawrence, A. D., & Barnard, P. J. (2008). Paying attention to social meaning: an FMRI study. Cereb Cortex, 18(8), 1876-1885. Taylor, P., Hobbs, J. N., Burroni, J., & Siegelmann, H. T. (2015). The global landscape of cognition: hierarchical aggregation as an organizational principle of human cortical networks and functions. Scientific reports, 5, 18112. Teffer, K., & Semendeferi, K. (2012). 9 Human prefrontal cortex: Evolution, development, and pathology. Progress in brain research, 195, 191. Todd, R. M., Cunningham, W. A., Anderson, A. K., & Thompson, E. (2012). Affect-biased attention as emotion regulation. Trends Cogn Sci, 16(7), 365-372. Todd, R. M., Talmi, D., Schmitz, T. W., Susskind, J., & Anderson, A. K. (2012). Psychophysical and neural evidence for emotion-enhanced perceptual vividness. J Neurosci, 32(33), 11201-11212. Tomita, H., Ohbayashi, M., Nakahara, K., Hasegawa, I., & Miyashita, Y. (1999). Top-down signal from prefrontal cortex in executive control of memory retrieval. Nature, 401(6754), 699-703. Turk-Browne, N. B. (2013). Functional interactions as big data in the human brain. Science, 342(6158), 580-584. Tusche, A., Smallwood, J., Bernhardt, B. C., & Singer, T. (2014). Classifying the wandering mind: Revealing the affective content of thoughts during task-free rest periods. Neuroimage, 97, 107-116. van den Heuvel, M. P., & Hulshoff Pol, H. E. (2010). Exploring the brain network: a review on resting-state fMRI functional connectivity. Eur Neuropsychopharmacol, 20(8), 519-534. van den Heuvel, M. P., & Sporns, O. (2011). Rich-club organization of the human connectome. The Journal of neuroscience : the official journal of the Society for Neuroscience, 31(44), 15775-15786. van den Heuvel, M. P., & Sporns, O. (2013). Network hubs in the human brain. Trends Cogn Sci, 17(12), 683-696. 156  van den Heuvel, M. P., Sporns, O., Collin, G., Scheewe, T., Mandl, R. C., Cahn, W., et al. (2013). Abnormal rich club organization and functional brain dynamics in schizophrenia. JAMA Psychiatry, 70(8), 783-792. Van Dijk, K. R., Hedden, T., Venkataraman, A., Evans, K. C., Lazar, S. W., & Buckner, R. L. (2010). Intrinsic functional connectivity as a tool for human connectomics: theory, properties, and optimization. J Neurophysiol, 103(1), 297-321. Van Essen, D. C. (2005). A population-average, landmark-and surface-based (PALS) atlas of human cerebral cortex. Neuroimage, 28(3), 635-662. Van Essen, D. C., & Dierker, D. L. (2007). Surface-based and probabilistic atlases of primate cerebral cortex. Neuron, 56(2), 209-225. Van Essen, D. C., Drury, H. A., Dickson, J., Harwell, J., Hanlon, D., & Anderson, C. H. (2001). An integrated software suite for surface-based analyses of cerebral cortex. Journal of the American Medical Informatics Association, 8(5), 443-459. Vanhaudenhuyse, A., Demertzi, A., Schabus, M., Noirhomme, Q., Bredart, S., Boly, M., et al. (2011). Two distinct neuronal networks mediate the awareness of environment and of self. J Cogn Neurosci, 23(3), 570-578. Verstynen, T. D. (2014). The organization and dynamics of corticostriatal pathways link the medial orbitofrontal cortex to future behavioral responses. Journal of Neurophysiology, 112(10), 2457-2469. Vincent, J. L., Kahn, I., Snyder, A. Z., Raichle, M. E., & Buckner, R. L. (2008). Evidence for a frontoparietal control system revealed by intrinsic functional connectivity. J Neurophysiol, 100(6), 3328-3342. Vincent, J. L., Patel, G. H., Fox, M. D., Snyder, A. Z., Baker, J. T., Van Essen, D. C., et al. (2007). Intrinsic functional architecture in the anaesthetized monkey brain. Nature, 447(7140), 83-86. Vogt, B. A. (2005). Pain and emotion interactions in subregions of the cingulate gyrus. Nat Rev Neurosci, 6(7), 533-544. Vuilleumier, P. (2005). How brains beware: neural mechanisms of emotional attention. Trends Cogn Sci, 9(12), 585-594. 157  Wagner, D. D., Haxby, J. V., & Heatherton, T. F. (2012). The Representation of Self and Person Knowledge in the Medial Prefrontal Cortex. Wiley Interdiscip Rev Cogn Sci, 3(4), 451-470. Wang, D., Buckner, R. L., Fox, M. D., Holt, D. J., Holmes, A. J., Stoecklein, S., et al. (2015). Parcellating cortical functional networks in individuals. Nature neuroscience, 18(12), 1853-1860. Watts, D. J., & Strogatz, S. H. (1998). Collective dynamics of 'small-world' networks. Nature, 393(6684), 440-442. Weissman, D. H., Roberts, K. C., Visscher, K. M., & Woldorff, M. G. (2006). The neural bases of momentary lapses in attention. Nat Neurosci, 9(7), 971-978. Weston, J., Elisseeff, A., BakIr, G., & Sinz, F. (2005). SPIDER: object-orientated machine learning library. URL http://www. kyb. tuebingen. mpg. de/bs/people/spider. Whitfield-Gabrieli, S., & Nieto-Castanon, A. (2012). Conn: a functional connectivity toolbox for correlated and anticorrelated brain networks. Brain connectivity, 2(3), 125-141. Wig, G. S., Schlaggar, B. L., & Petersen, S. E. (2011). Concepts and principles in the analysis of brain networks. Annals of the New York Academy of Sciences, 1224, 126-146. Woolgar, A., Hampshire, A., Thompson, R., & Duncan, J. (2011). Adaptive coding of task-relevant information in human frontoparietal cortex. J Neurosci, 31(41), 14592-14599. Wotruba, D., Michels, L., Buechler, R., Metzler, S., Theodoridou, A., Gerstenberg, M., et al. (2013). Aberrant Coupling Within and Across the Default Mode, Task-Positive, and Salience Network in Subjects at Risk for Psychosis. Schizophr Bull. Xue, S., Tang, Y. Y., & Posner, M. I. (2011). Short-term meditation increases network efficiency of the anterior cingulate cortex. Neuroreport, 22(12), 570-574. Yarkoni, T., Poldrack, R. A., Nichols, T. E., Van Essen, D. C., & Wager, T. D. (2011). Large-scale automated synthesis of human functional neuroimaging data. Nat Methods, 8(8), 665-670. Yeo, B. T., Krienen, F. M., Sepulcre, J., Sabuncu, M. R., Lashkari, D., Hollinshead, M., et al. (2011). The organization of the human cerebral cortex estimated by intrinsic functional connectivity. J Neurophysiol, 106(3), 1125-1165. Yeo, B. T., Tandi, J., & Chee, M. W. (2015). Functional connectivity during rested wakefulness predicts vulnerability to sleep deprivation. Neuroimage, 111, 147-158. 158  Zabelina, D. L., & Andrews-Hanna, J. R. (2016). Dynamic network interactions supporting internally-oriented cognition. Current opinion in neurobiology, 40, 86-93. Zalesky, A., Fornito, A., Cocchi, L., Gollo, L. L., & Breakspear, M. (2014). Time-resolved resting-state brain networks. Proceedings of the National Academy of Sciences of the United States of America, 111(28), 10341-10346. Zeidman, P., Mullally, S. L., & Maguire, E. A. (2015). Constructing, Perceiving, and Maintaining Scenes: Hippocampal Activity and Connectivity. Cerebral cortex, 25(10), 3836-3855.    159  APPENDICES Appendix A: Supplementary material for experiment 1 Table 1. Details for studies included in the analysis of effect sizes Study correlation r or z(r) Noise removal Data extracted from Golland et al. 2007  0.09 r GSR figure 4 Kelly et al. 2008 -0.89 r GSR results text Murphy et al. 2009 -0.72 r GSR average based on Table 1   0.18 r No GSR average based on Table 1 Chang & Glover 2009 -0.35 r GSR approximated from Figure 1  -0.25 r No GSR approximated from Figure 1 Van Dijk et al 2010 -0.24 z (r ) GSR results text   0.16 z (r ) No GSR results text Anderson et al. 2011a  0.05 z (r ) No GSR approximated from Figure 3 Fornito et al. 2012 -0.50 r GSR approximated from Figure S2  -0.40 r No GSR approximated from Figure S3 Lee et al. 2012 -0.74 r GSR results text 160  Chai et al. 2012 -0.20 z (r ) GSR approximated from Figure 5  -0.12 z (r ) No GSR approximated from Figure 5 Gao and Lin. 2012 -0.20 r GSR approximated from Figure 2 De Havas et al. 2012 -0.26 r GSR approximated from Figure 2 Josipovic et al. 2012 -0.16 r GSR approximated from Figure 4b Cole et al. 2014 -0.07 z(r) GSR data provided by authors Chai et al. 2014 -0.15 r No GSR approximated from Figure 2 & 4 Wotruba et al. 2014 -0.02 r No GSR figure 2 Holmes et al 2015 -0.18 z (r ) GSR data provided by authors Yeo et al. 2015 -0.50 r GSR figure 7   0.38 r No GSR figure 7 Spreng et al. 2016 -0.04 z (r ) GSR data provided by authors   0.08 z (r ) No GSR data provided by authors Amer et al. 2016 -0.06 z (r ) No GSR average based on data points in figure 2 Current study -0.06 z (r ) No GSR       161   Table 2. Correlations between self-report variables and DN-DAN functional connectivity   Cognitive State Variable Subsystem Rest Movie Art Shopping Evaluation Acceptance Attention Core  0.24 0.31 0.06    DM  0.34 0.04 -0.18    MTL  0.04 0.04 -0.03           Difficulty Core  -0.35 -0.11 0.21 -0.05 -0.10  DM  -0.37 -0.05 -0.11 -0.25 -0.02  MTL  -0.03 -0.03 0.09 0.02 -0.42         Familiarity Core  0.10 -0.06 -0.08    DM  0.41 -0.06 -0.24    MTL  -0.25 -0.20 -0.25           Enjoyment Core  0.19 -0.02 -0.22    DM  0.29 -0.01 -0.19    MTL  -0.02 -0.02 -0.05   Note: For exploratory purposes only we report the correlation values between individual differences in self-reported data and the strength of FC between the DAN and each DN 162  subsystem. Each variable was rated on a 7-point scale, from 1=low to 7=high. Given our small sample size, these values should be interpreted with caution.    163   Figure 1. Mean functional connectivity between the DAN and each DN subsystem for each of the 24 participants. Not a single participant exhibited a linear change in mean between-network FC across contexts, ruling out the possibility that a general effect (e.g., fatigue) can explain the effect of context. In fact, participants show varying patterns of mean between-network FC across 164  contexts; reliable group-level effects are only observed in terms of the pattern across specific between-network nodes-to-node connections (see Figure 4).          Figure 2. ROIs for the DAN and three DN subsystems. The ROIs were created by Yeo and colleagues (Krienen, Yeo, & Buckner, 2014; Yeo, Tandi, & Chee, 2015) based on their 17-network parcellation (Yeo et al., 2011). The DAN (green) included the frontal eye fields (FEFs), anterior intraparietal sulcus/superior parietal lobule (aIPS/SPL), ventral precentral cortex (PrCv), and anterior middle temporal region (aMT). The DN core subsystem (yellow) included the rostromedial prefrontal cortex (RMPFC), posterior cingulate cortex (PCC), posterior inferior parietal lobule (pIPL), superior frontal sulcus (SFS), and right rostral superior temporal sulcus (rSTS). The DN dorsomedial prefrontal subsystem (red) included the dorsomedial prefrontal cortex (DMPFC), left temporoparietal junction (TPJ), temporopolar cortex (TP), lateral temporal cortex (LTC), inferior frontal gyrus (IFG), and left posterior dorsolateral prefrontal cortex (pDLPFC). The DN medial temporal lobe subsystem (blue) included the medial temporal lobe (MTL; specifically the parahippocampal gyrus), retrosplenial cortex/ventral posterior cingulate cortex (RSC/vPCC), and ventral posterior inferior parietal lobule (vpIPL).  165    Figure 3. Accuracy in distinguishing each pair of cognitive contexts using a linear SVM classifier. Error bars reflect between-subject SEM.    Figure 4. Accuracy in distinguishing each pair of cognitive contexts using a neural network classifier and 4-fold cross validation. Error bars reflect between-fold SEM. 166   Figure 5. Seed-based connectivity analyses showing negative connectivity with DAN regions (Z > 3.1, p < .05 FWE corrected for cluster extent), with the borders of each DN subsystem highlighted. Data for right hemisphere. FEF, Frontal eye fields; aIPS/SPL, anterior intraparietal sulcus/superior parietal lobule; PrCv, ventral precentral cortex; aMT, anterior middle temporal region. Color bar represents t-values.       Figure 6. Mean within- and across-context similarity of anticorrelations. In this case, across-context similarity was based on comparing connectivity in adjacent conditions, thus ensuring that separation in time was not driving the lower across-context similarity. Across-context similarity was again low in this case indicating that the results reflect a true reconfiguration of DN-DAN 167  interactions across contexts. DN, entire default network, DM, dorsomedial subsystem; MT, medial temporal subsystem. Error bars reflect within-subject SEM (Loftus & Masson, 1994).   168    Figure 7. Positive and negative connectivity for each DAN seed region, for each context. Data for right hemisphere. Negative FC flexibly increased and decreased in different cognitive contexts relative to rest. For illustration purposes, we use a slightly liberal threshold to show the full extent of positively and negatively correlated voxels in each context (Z > 2.57, p < .05 FDR cluster corrected). FEF, Frontal eye fields; aIPS/SPL, anterior intraparietal sulcus/superior parietal lobule; PrCv, ventral precentral cortex; aMT, anterior middle temporal region.   169      Figure 8. ROIs for the frontoparietal control network. The ROIs were created by Yeo and colleagues (Krienen et al., 2014; Yeo et al., 2015) based on their 17-network parcellation (Yeo et al., 2011) and included the rostrolateral prefrontal cortex (RLPFC), posterior dorsolateral prefrontal cortex (pDLPFC), anterior inferior parietal lobule (aIPL), posterior dorsomedial prefrontal cortex (pDMPFC), and middle temporal gyrus (MTG).    170  Results We attempted to limit the possibility that motion may influence FC values through several preprocessing steps (i.e., regressing out outlier time-points, as well as timeseries signals from the white matter and ventricles). Nevertheless, motion may still have an impact, and we thus conducted control analyses to ensure that motion could not explain the effect of context on FC. We examined total motion (TM; the sum of motion across six movement dimensions at each time-point) and framewise displacement (FD; the difference in motion between adjacent time-points). Both TM and FD were very similar across contexts. On average, participants exhibited less than 1 mm in TM within each context (rest: 0.88 mm; movie viewing: 0.88 mm; artwork analysis: 0.68 mm; shopping task: 0.75 mm; evaluation-based introspection: 0.99 mm; acceptance-based introspection: 1.04 mm). TM did not differ across contexts [F(5, 110) = 1.82, p = .12]. Average FD was also very small within each context (rest: 0.17 mm; movie viewing: 0.14 mm; artwork analysis: 0.14 mm; shopping task: 0.15 mm; evaluation-based introspection: 0.16 mm; acceptance-based introspection: 0.19 mm). FD did show a difference across contexts [F(5, 110) = 4.64, p = .001] due to less displacement during movie viewing and artwork analysis relative to rest (ps < .05). However, the actual differences in movement between conditions were extremely small. Next, we examined whether the amount of variation in movement across contexts was correlated with the amount of variation in DAN-DN FC strength. Notably, in all cases there was no relationship or a negative relationship, indicating that those participants that showed the greatest change in DN-DAN FC across contexts were the ones that exhibited the least amount of movement across contexts. After removing one outlier (> 2 SD from mean), we found no relationship or a negative relationship between FC variability and TM variability (DAN-Core left: r = -.36, p = .09; DAN-Core right: r = -.39, p = .066; DAN-dorsomedial prefrontal subsystem left: r = -.52, p = .01; DAN-dorsomedial prefrontal subsystem right: r = -.12, p = .58; DAN-MTL subsystem left: r = -.03, p = .89; DAN-MTL subsystem right: r = -.23, p = .29), and no relationship or a negative relationship between FC variability and FD variability  (DAN-Core left: r = -.49, p = .018; DAN-Core right: r = -.06, p = .78; DAN-dorsomedial prefrontal subsystem left: r = .04, p = .86; DAN-dorsomedial prefrontal subsystem right: r = .01, p = .96; DAN-MTL subsystem left: r = -.15, p = .49; DAN-MTL subsystem right: r = -.18, p = .41). Thus, motion was generally associated with less influence of context on FC strength. To probe deeper, we looked more specifically at whether the change (Δ) in motion from one context 171  to the next was correlated with change in functional connectivity strength from one context to the next. Of 30 correlation analyses performed (3 subsystems x 2 hemispheres x 5 context changes), there was no significant associations between Δ TM and Δ FC at a liberal p < .05 uncorrected threshold. Furthermore, of 30 correlation analyses performed, there was only one significant associations between Δ FD and Δ FC at a liberal p < .05 uncorrected threshold (evaluation-based introspection to acceptance-based introspection, DAN-MTL subsystem right: r = .52, p = .01). These analyses provide robust evidence that the effect of context on functional connectivity was not driven by motion.   172  Appendix B: Supplementary material for experiment 2   Figure 1. Hierarchical clustering of FPCN (right hemisphere). (A) Clustering based on within-FPCN functional connections. (B) Clustering based on extrinsic (DN and DAN) functional connections.     Figure 2. Classifier accuracy in distinguishing aFPCN and pFPCN FC patterns. (A) Classification accuracy based on FC with the DN and DAN using an ANOVA kernel. (B) Classification accuracy based on FC with just the DN using a linear kernel. (C) Classification accuracy based on FC with just the DAN using a linear kernel. Error bars reflect between subject SEM.   173    Figure 3. Seed maps. FPCN seed maps during rest (Z > 3.1, p < .05 FWE cluster corrected). Default network borders (yellow) and dorsal attention network borders (green) were demarcated based on Yeo et al.'s (2011) 17-network parcellation.   174   Figure 3 continued.     Figure 4. FPCN FC patterns with the dorsomedial prefrontal and medial temporal lobe subsystems of the DN. Error bars reflect between subject SEM.   175    Figure 5. Mean node selectivity. The “selectivity index” reflects the extent to which aFPCN nodes are preferentially coupled with DN nodes compared to DAN nodes, and the extent to which pFPCN nodes are preferentially coupled with DAN nodes compared to DN nodes. The selectivity index is stronger for the aFPCN than the pFPCN. Error bars reflect between subject SEM.      Figure 6. The “selectivity index” varies in strength across conditions for both FPCN subsystems. Error bars reflect between subject SEM.     176   Figure 7. Variability of FC with the DN and DAN across conditions for each FPCN node. Values reflect mean variability (SD) of FC across conditions between each FPCN node and DN and DAN nodes. Error bars reflect between subject SEM.      Figure 8. Neurosynth forward inference meta-analyses. Forward inference maps reflect z-scores corresponding to the probability of activation in a given region if a study uses a particular term (P(Activation|Term)). Studies using the term “perception”, or “action”, or consistently engage the pFPCN, but show very little engagement of the aFPCN.      177  Results Control for node distance  In many cases aFPCN nodes are spatially proximate to DN nodes, whereas pFPCN nodes are spatially proximate to DAN nodes. As such, differential coupling patterns could result from spatial smoothing which may increase coupling strength between adjacent regions. We conducted a control analysis to rule out the possibility that this could explain our results. We computed the average correlation between the rostrolateral prefrontal cortex (RLPFC)―a node within the aFPCN―and several DN regions including the posterior cingulate cortex (PCC), posterior inferior parietal lobule (pIPL), superior frontal sulcus (SFS), and lateral temporal cortex (LTC), and compared this to the average correlation between the inferior frontal sulcus (IFS)―a node within the pFPCN―and these same regions. The IFS is spatially closer to these DN regions, so it should exhibit stronger FC with these regions if our results are mainly driven by spatial proximity. However, this was not the case. RLPFC coupling with these DN regions [mean z(r) = .33] was much stronger than IFS coupling with these regions [mean z(r) = .05], a difference that was statistically significant (p < .001). We also computed the average correlation between the middle temporal gyrus (MTG)―a node within the aFPCN―and several DAN regions including the frontal eye fields and ventral precentral cortex, and compared this to the average correlation between the posterior middle temporal gyrus (pMTG)―a node within the pFPCN―and these same regions. In this case, the MTG is spatially closer to these DAN regions, so it should exhibit stronger FC with these regions if our results are mainly driven by spatial proximity. However, this was not the case. MTG coupling with these DAN regions [mean z(r) = -.01] was much weaker than pMTG coupling with these regions [mean z(r) = .11], a difference that was statistically significant (p < .001). Thus, the aFPCN exhibits stronger coupling with the DN and the pFPCN exhibits stronger coupling with the DAN even when spatial proximity of nodes cannot be a factor.    Control for motion It is important to rule out the possibility that the relationship between dynamic FC interactions and network clustering could be explained by participant motion. We examined this in our primary data set. We examined total motion and framewise displacement. We computed the 178  average amount of motion in each window, just as with between-network FC, and then computed the correlation between changes across time in motion and changes across time in between-network FC and clustering. These values were Fisher-transformed and submitted to a one-sample t-test to assess statistical significance at the group level, with p = .004 corresponding to p < .05 Bonferroni corrected for multiple comparisons. Notably, within each of the six contexts, temporal variation in the strength of between-network FC and temporal variation in the strength of clustering were uncorrelated with temporal variation in participant motion. We found no significant relationships at the group level in any context between motion and between-network FC [all: |z(r)'s| < .09, p's > .05, FWE corrected], or between motion and mean weighted clustering [all: |z(r)'s| < .12, p's > .05, FWE corrected]. There was also no evidence of systematic relationships at the level of individual participants. For between-network FC we found that 19 out of 284 correlations (~ 7%) were significantly positive at p < .05, Bonferroni corrected, and 22 out of 284 correlations (~ 8%) were significantly negative at p < .05, Bonferroni corrected. For clustering, we found that 33 out of 284 correlations (~ 12%) were significantly positive at p < .05, Bonferroni corrected, and 39 out of 284 correlations (~ 14%) were significantly negative at p < .05, Bonferroni corrected. Thus, while some participants did show significant correlations with motion in some contexts, this was rare, and the correlations were not systematically positive or negative. Thus, the observed time-dependent relationships between network interactions and clustering strength cannot be explained by participant motion.         

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.24.1-0354463/manifest

Comment

Related Items